      ‚  ‚  '  N  ™                    /hY               D     )ÑÒ€     Ò»Õ¬    »Õ;×¬    ;×ÏÙ`    ÏÙ©ÛØ    ©Û•Ýl    •ÝßÞ¬    ßÞà\	    à5ã
    5ã+ä8    +ä›éd    ›é/ì|    /ìWîô    Wî£ðT    £ðµóR     µó5õ:     5õ/4â	                              b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   W  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       T        £   «          «   ³          ³   »         »   Ã          ¼   Ä          Ä   Ì         Ë   Ó          Ì   Ô          Ô   Ü          Ü   ä          ä   ì         ì   ô          ì   ô          ô   ü          ü                                                         $         $  ,        ,  4         ,  4         4  <         <  D         D  L        L  T         L  T         T  \         \  d         d  l        l  t         l  t         t  |         |  „         „  Œ        Œ  ”         Œ  ”         ”  œ         œ  ¤         ¤  ¬        ¬  ´         ¬  ´         ´  ¼         ¼  Ä         Ä  Ì        Ì  Ô         Ì  Ô         Ô  Ü         Ü  ä         ä  ì        ì  ô         ì  ô         ô  ü         ü                                                        $         $  ,        ,  4         ,  4         4  <         <  D         D  L        L  T         L  T         T  \         \  d         d  l        l  t         l  t         t  |         |  „         „  Œ        Œ  ”         Œ  ”         ”  œ         œ  ¤         ¤  ¬        ¬  ´         ¬  ´         ´  ¼         ¼  Ä                             b      J      w   �      operation.0     copy_u64              5   :   5   R   9   2   9      9   2   9   "   9   B   9   Z   =   b   A      A   š   I   *   I   R   M   B   reason  LocalCopy       value   13      bytes   512     pattern Contiguous      invocations     1       uniformInvocations      true    arguments       [10, 4]                               b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1               5   e          6   V          V   v         e   •          v   –         •   Å                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   —   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       $         +   3          3   ;         5   7          ;   A          A   E          E   M          M   Q         S   [          T   V                              b      ‚      w   �      operation.1     static_copy_u32       5   :   5   R   9   2   9      9   2   9      9   B   9   Z   =   b   A      A   š   I   *   I   R   M   "   reason  LocalCopy       value   16      bytes   32      pattern Contiguous      invocations     1       uniformInvocations      true    arguments       [8]                                   b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]           	   	                    b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]         
     €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        8   ¸          ®   ¶         Æ   Î         Î   Ö         Ö   Þ         ë   ó         ô   ú                                                   !  )        '  /        <  >        >  @        @  H        J  R        [  c        “  ™        ™  ¡        £  «        «  ³        ³  »        »  Ã                             b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       &         b      J      7         operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1                              b   ò      
     B     operation.2        \               "              Â
            Š            b)            **            ò*            º+            ‚,            J-     	       .     
       Ú.            ¢/            j0            21            ú1            Â2            Š3            R4            5            â5            ª6            r7            :8            9            Ê9            ’:            Z;            "<            ê<            ²=            z>            B?             
@     !       Ò@     "       šA     #       bB     $       *C     %       òC     &       ºD     '       ‚E     (       JF     )       G     *       ÚG     +       ¢H     ,       jI     -       2J     .       úJ     /       ÂK     0       ŠL     1       RM     2       N     3       âN     4       ªO     5       rP     6       :Q     7       R     8       ÊR     9       ’S     :       ZT     ;       "U     <       êU     =       ²V     >       zW     ?       BX     @       
Y     A       òv     B       ºw     C       ’•     D       Z–     E       "—     F       ê—     G       ²˜     H       z™     I       Bš     J       
›     K       Ò›     L       ò›     M       b     N       *     O       ò     P       º     Q       ‚     R       J     S            T       Ú     U       ¢     V       j     W       2	     X       ú	     Y       Â
     Z       Š     [       R     \            ]       â     ^       ª     _       r     `       :     a            b       Ê     c       ’     d       Z     e       "     f       ê     g       ²     h       z     i       B     j       
     k       Ò     l       š     m       b     n       *     o       ò     p       º     q       ‚     r       J     s            t       Ú     u       ¢      v       j!     w       2"     x       ú"     y       Â#     z       Š$     {       R%     |       &     }       â&     ~       ª'            r(     €       :)     �       G     ‚       ÊG     ƒ       úf     „       Âg     …       Šh     †       Ri     ‡       j     ˆ       âj     ‰       ªk     Š       rl     ‹       :m     Œ       n     �       Ên     Ž       ’o     �       Zp     �       "q     ‘       êq     ’       ²r     “       zs     ”       Bt     •       
u     –       Òu     —       šv     ˜       bw     ™       *x     š       òx     ›       ºy     œ       ‚z     �       J{     ž       |     Ÿ       Ú|             ¢}     ¡       j~     ¢       2     £       ú     ¤       Â€     ¥       Š�     ¦       R‚     §       ƒ     ¨       âƒ     ©       ª„     ª       r…     «       :†     ¬       ‡     ­       Ê‡     ®       ’ˆ     ¯       Z‰     °       "Š     ±       êŠ     ²       ²‹     ³       zŒ     ´       B�     µ       
Ž     ¶       ÒŽ     ·       š�     ¸       b�     ¹       *‘     º       ò‘     »       º’     ¼       ‚“     ½       J”     ¾       •     ¿       Ú•     À       ¢–     Á       âµ     Â       ª¶     Ã       êÕ     Ä       ²Ö     Å       z×     Æ       BØ     Ç       
Ù     È       ÒÙ     É       šÚ     Ê       bÛ     Ë       *Ü     Ì       òÜ     Í       ºÝ     Î       ‚Þ     Ï       Jß     Ð       à     Ñ       Úà     Ò       ¢á     Ó       jâ     Ô       2ã     Õ       úã     Ö       Âä     ×       Šå     Ø       Ræ     Ù       ç     Ú       âç     Û       ªè     Ü       ré     Ý       :ê     Þ       ë     ß       Êë     à       ’ì     á       Zí     â       "î     ã       êî     ä       ²ï     å       zð     æ       Bñ     ç       
ò     è       Òò     é       šó     ê       bô     ë       *õ     ì       òõ     í       ºö     î       ‚÷     ï       Jø     ð       ù     ñ       Úù     ò       ¢ú     ó       jû     ô       2ü     õ       úü     ö       Âý     ÷       Šþ     ø       Rÿ     ù            ú       â     û       ª    ü       r    ý       :    þ           ÿ       Ê           ’          *#          ò#          Ê	           ’
           Z           "           ê           ²     	      z     
      B           
           Ò           š           b           *           ò           º           ‚           J                      Ú           ¢           j           2           ú           Â           Š           R                      â           ª            r      !      :!     "      "     #      Ê"     $      ’#     %      Z$     &      "%     '      ê%     (      ²&     )      z'     *      B(     +      
)     ,      Ò)     -      š*     .      b+     /      *,     0      ò,     1      º-     2      ‚.     3      J/     4      0     5      Ú0     6      ¢1     7      j2     8      23     9      ú3     :      Â4     ;      Š5     <      R6     =      7     >      â7     ?      ª8     @      r9     A      
W     B      ÒW     C      zu     D      Bv     E      
w     F      Òw     G      šx     H      by     I      *z     J      òz     K      º{     L      ‚|     M      J}     N      ~     O      Ú~     P      ¢     Q      j€     R      2�     S      ú�     T      Â‚     U      Šƒ     V      R„     W      …     X      â…     Y      ª†     Z      r‡     [      :ˆ     \      ‰     ]      Ê‰     ^      ’Š     _      Z‹     `      "Œ     a      êŒ     b      ²�     c      zŽ     d      B�     e      
�     f      Ò�     g      š‘     h      b’     i      *“     j      ò“     k      º”     l      ‚•     m      J–     n      —     o      Ú—     p      ¢˜     q      j™     r      2š     s      úš     t      Â›     u      Šœ     v      R�     w      ž     x      âž     y      ªŸ     z      r      {      :¡     |      ¢     }      Ê¢     ~      ’£           Z¤     €      "¥     �      ºÂ     ‚      ‚Ã     ƒ      á     „      âá     …      ªâ     †      rã     ‡      :ä     ˆ      å     ‰      Êå     Š      ’æ     ‹      Zç     Œ      "è     �      êè     Ž      ²é     �      zê     �      Bë     ‘      
ì     ’      Òì     “      ší     ”      bî     •      *ï     –      òï     —      ºð     ˜      ‚ñ     ™      Jò     š      ó     ›      Úó     œ      ¢ô     �      jõ     ž      2ö     Ÿ      úö            Â÷     ¡      Šø     ¢      Rù     £      ú     ¤      âú     ¥      ªû     ¦      rü     §      :ý     ¨      þ     ©      Êþ     ª      ’ÿ     «      Z     ¬      "    ­      ê    ®      ²    ¯      z    °      B    ±      
    ²      Ò    ³      š    ´      b    µ      *    ¶      ò    ·      º	    ¸      ‚
    ¹      J    º          »      Ú    ¼      ¢    ½      j    ¾      2    ¿      ú    À      Â    Á      J.    Â      /    Ã      šL    Ä      bM    Å      *N    Æ      òN    Ç      ºO    È      ‚P    É      JQ    Ê      R    Ë      ÚR    Ì      ¢S    Í      jT    Î      2U    Ï      úU    Ð      ÂV    Ñ      ŠW    Ò      RX    Ó      Y    Ô      âY    Õ      ªZ    Ö      r[    ×      :\    Ø      ]    Ù      Ê]    Ú      ’^    Û      Z_    Ü      "`    Ý      ê`    Þ      ²a    ß      zb    à      Bc    á      
d    â      Òd    ã      še    ä      bf    å      *g    æ      òg    ç      ºh    è      ‚i    é      Jj    ê      k    ë      Úk    ì      ¢l    í      jm    î      2n    ï      ún    ð      Âo    ñ      Šp    ò      Rq    ó      r    ô      âr    õ      ªs    ö      rt    ÷      :u    ø      v    ù      Êv    ú      ’w    û      Zx    ü      "y    ý      êy    þ      ²z    ÿ      z{           B|          Ê™          ’š          ¸          â¸          ª¹          rº          :»          ¼    	      Ê¼    
      ’½          Z¾          "¿          ê¿          ²À          zÁ          BÂ          
Ã          ÒÃ          šÄ          bÅ          *Æ          òÆ          ºÇ          ‚È          JÉ          Ê          ÚÊ          ¢Ë          jÌ          2Í          úÍ           ÂÎ    !      ŠÏ    "      RÐ    #      Ñ    $      âÑ    %      ªÒ    &      rÓ    '      :Ô    (      Õ    )      ÊÕ    *      ’Ö    +      Z×    ,      "Ø    -      êØ    .      ²Ù    /      zÚ    0      BÛ    1      
Ü    2      ÒÜ    3      šÝ    4      bÞ    5      *ß    6      òß    7      ºà    8      ‚á    9      Jâ    :      ã    ;      Úã    <      ¢ä    =      jå    >      2æ    ?      úæ    @      Âç    A      J    B          C      š#    D      b$    E      *%    F      ò%    G      º&    H      ‚'    I      J(    J      )    K      Ú)    L      ¢*    M      j+    N      2,    O      ú,    P      Â-    Q      Š.    R      R/    S      0    T      â0    U      ª1    V      r2    W      :3    X      4    Y      Ê4    Z      ’5    [      Z6    \      "7    ]      ê7    ^      ²8    _      z9    `      B:    a      
;    b      Ò;    c      š<    d      b=    e      *>    f      ò>    g      º?    h      ‚@    i      JA    j      B    k      ÚB    l      ¢C    m      jD    n      2E    o      úE    p      ÂF    q      ŠG    r      RH    s      I    t      âI    u      ªJ    v      rK    w      :L    x      M    y      ÊM    z      ’N    {      ZO    |      "P    }      êP    ~      ²Q          zR    €      BS    �      š      ‚      b     ƒ      ê     „      ²     …      z      †      B!     ‡      
"     ˆ      Ò"     ‰      š#     Š      b$     ‹      *%     Œ      ò%     �      º&     Ž      ‚'     �      J(     �      )     ‘      Ú)     ’      ¢*     “      j+     ”      2,     •      ú,     –      Â-     —      Š.     ˜      R/     ™      0     š      â0     ›      ª1     œ      r2     �      :3     ž      4     Ÿ      Ê4            ’5     ¡      Z6     ¢      "7     £      ê7     ¤      ²8     ¥      z9     ¦      B:     §      
;     ¨      Ò;     ©      š<     ª      b=     «      *>     ¬      ò>     ­      º?     ®      ‚@     ¯      JA     °      B     ±      ÚB     ²      ¢C     ³      jD     ´      2E     µ      úE     ¶      ÂF     ·      ŠG     ¸      RH     ¹      I     º      âI     »      ªJ     ¼      rK     ½      :L     ¾      M     ¿      ÊM     À      ’N     Á      l     Â      âl     Ã      jŠ     Ä      2‹     Å      ú‹     Æ      ÂŒ     Ç      Š�     È      RŽ     É      �     Ê      â�     Ë      ª�     Ì      r‘     Í      :’     Î      “     Ï      Ê“     Ð      ’”     Ñ      Z•     Ò      "–     Ó      ê–     Ô      ²—     Õ      z˜     Ö      B™     ×      
š     Ø      Òš     Ù      š›     Ú      bœ     Û      *�     Ü      ò�     Ý      ºž     Þ      ‚Ÿ     ß      J      à      ¡     á      Ú¡     â      ¢¢     ã      j£     ä      2¤     å      ú¤     æ      Â¥     ç      Š¦     è      R§     é      ¨     ê      â¨     ë      ª©     ì      rª     í      :«     î      ¬     ï      Ê¬     ð      ’­     ñ      Z®     ò      "¯     ó      ê¯     ô      ²°     õ      z±     ö      B²     ÷      
³     ø      Ò³     ù      š´     ú      bµ     û      *¶     ü      ò¶     ý      º·     þ      ‚¸     ÿ      J¹            º           Š×           RØ           Êõ           ’ö           Z÷           "ø           êø           ²ù     	      zú     
      Bû           
ü           Òü           šý           bþ           *ÿ           òÿ           º           ‚          J                    Ú          ¢          j          2          ú          Â          Š          R	          
          â
          ª           r    !      :    "          #      Ê    $      ’    %      Z    &      "    '      ê    (      ²    )      z    *      B    +      
    ,      Ò    -      š    .      b    /      *    0      ò    1      º    2      ‚    3      J    4          5      Ú    6      ¢    7      j    8      2    9      ú    :      Â     ;      Š!    <      R"    =      #    >      â#    ?      ª$    @      r%    A      êB    B      ²C    C      *a    D      òa    E      ºb    F      ‚c    G      Jd    H      e    I      Úe    J      ¢f    K      jg    L      2h    M      úh    N      Âi    O      Šj    P      Rk    Q      l    R      âl    S      ªm    T      rn    U      :o    V      p    W      Êp    X      ’q    Y      Zr    Z      "s    [      ês    \      ²t    ]      zu    ^      Bv    _      
w    `      Òw    a      šx    b      by    c      *z    d      òz    e      º{    f      ‚|    g      J}    h      ~    i      Ú~    j      ¢    k      j€    l      2�    m      ú�    n      Â‚    o      Šƒ    p      R„    q      …    r      â…    s      ª†    t      r‡    u      :ˆ    v      ‰    w      Ê‰    x      ’Š    y      Z‹    z      "Œ    {      êŒ    |      ²�    }      zŽ    ~      B�          
�    €      Ò�    �      :®    ‚      ·    ƒ      zÔ    „      RÝ    …      Þ    †      âÞ    ‡      ªß    ˆ      rà    ‰      :á    Š      â    ‹      Êâ    Œ      ’ã    �      Zä    Ž      "å    �      êå    �      ²æ    ‘      zç    ’      Bè    “      
é    ”      Òé    •      šê    –      bë    —      *ì    ˜      òì    ™      ºí    š      ‚î    ›      Jï    œ      ð    �      Úð    ž      ¢ñ    Ÿ      jò           2ó    ¡      úó    ¢      Âô    £      Šõ    ¤      Rö    ¥      ÷    ¦      â÷    §      ªø    ¨      rù    ©      :ú    ª      û    «      Êû    ¬      ’ü    ­      Zý    ®      "þ    ¯      êþ    °      ²ÿ    ±      z     ²      B    ³      
    ´      Ò    µ      š    ¶      b    ·      *    ¸      ò    ¹      º    º      ‚    »      J    ¼      	    ½      Ú	    ¾      ¢
    ¿      j    À      2    Á      š)    Â      r2    Ã      BQ    Ä      Z    Å      âZ    Æ      ª[    Ç      r\    È      :]    É      ^    Ê      Ê^    Ë      ’_    Ì      Z`    Í      "a    Î      êa    Ï      ²b    Ð      zc    Ñ      Bd    Ò      
e    Ó      Òe    Ô      šf    Õ      bg    Ö      *h    ×      òh    Ø      ºi    Ù      ‚j    Ú      Jk    Û      l    Ü      Úl    Ý      ¢m    Þ      jn    ß      2o    à      úo    á      Âp    â      Šq    ã      Rr    ä      s    å      âs    æ      ªt    ç      ru    è      :v    é      w    ê      Êw    ë      ’x    ì      Zy    í      "z    î      êz    ï      ²{    ð      z|    ñ      B}    ò      
~    ó      Ò~    ô      š    õ      b€    ö      *�    ÷      ò�    ø      º‚    ù      ‚ƒ    ú      J„    û      …    ü      Ú…    ý      ¢†    þ      j‡    ÿ      2ˆ           úˆ          Ò‘          ªš          ‚£          Z¬          "­          ê­          ²®          z¯    	      B°    
      
±          Ò±          š²          b³          *´          ò´          ºµ          ‚¶          J·          ¸          Ú¸          ¢¹          jº          2»          ú»          Â¼          Š½          R¾          ¿          â¿          ªÀ          rÁ           :Â    !      Ã    "      ÊÃ    #      ’Ä    $      ZÅ    %      "Æ    &      êÆ    '      ²Ç    (      zÈ    )      BÉ    *      
Ê    +      ÒÊ    ,      šË    -      bÌ    .      *Í    /      òÍ    0      ºÎ    1      ‚Ï    2      JÐ    3      Ñ    4      ÚÑ    5      ¢Ò    6      jÓ    7      2Ô    8      úÔ    9      ÂÕ    :      ŠÖ    ;      R×    <      Ø    =      âØ    >      ªÙ    ?      rÚ    @      :Û    A      ä    B      êì    C      Âõ    D      šþ    E      bÿ    F      *     G      ò     H      º    I      ‚    J      J    K          L      Ú    M      ¢    N      j    O      2    P      ú    Q      Â    R      Š	    S      R
    T          U      â    V      ª    W      r    X      :    Y          Z      Ê    [      ’    \      Z    ]      "    ^      ê    _      ²    `      z    a      B    b      
    c      Ò    d      š    e      b    f      *    g      ò    h      º    i      ‚    j      J    k          l      Ú    m      ¢    n      j    o      2     p      ú     q      Â!    r      Š"    s      R#    t      $    u      â$    v      ª%    w      r&    x      :'    y      (    z      Ê(    {      ’)    |      Z*    }      "+    ~      ê+          ²,    €      z-    �      R6    ‚      *?    ƒ      H    „      ÚP    …      ¢Q    †      jR    ‡      2S    ˆ      úS    ‰      ÂT    Š      ŠU    ‹      RV    Œ      W    �      âW    Ž      ªX    �      rY    �      :Z    ‘      [    ’      Ê[    “      ’\    ”      Z]    •      "^    –      ê^    —      ²_    ˜      z`    ™      Ba    š      
b    ›      Òb    œ      šc    �      bd    ž      *e    Ÿ      òe           ºf    ¡      ‚g    ¢      Jh    £      i    ¤      Úi    ¥      ¢j    ¦      jk    §      2l    ¨      úl    ©      Âm    ª      Šn    «      Ro    ¬      p    ­      âp    ®      ªq    ¯      rr    °      :s    ±      t    ²      Êt    ³      ’u    ´      Zv    µ      "w    ¶      êw    ·      ²x    ¸      zy    ¹      Bz    º      
{    »      Ò{    ¼      š|    ½      b}    ¾      *~    ¿      ò~    À      º    Á      ’ˆ    Â      j‘    Ã      Bš    Ä      £    Å      â£    Æ      ª¤    Ç      r¥    È      :¦    É      §    Ê      Ê§    Ë      ’¨    Ì      Z©    Í      "ª    Î      êª    Ï      ²«    Ð      z¬    Ñ      B­    Ò      
®    Ó      Ò®    Ô      š¯    Õ      b°    Ö      *±    ×      ò±    Ø      º²    Ù      ‚³    Ú      J´    Û      µ    Ü      Úµ    Ý      ¢¶    Þ      j·    ß      2¸    à      ú¸    á      Â¹    â      Šº    ã      R»    ä      ¼    å      â¼    æ      ª½    ç      r¾    è      :¿    é      À    ê      ÊÀ    ë      ’Á    ì      ZÂ    í      "Ã    î      êÃ    ï      ²Ä    ð      zÅ    ñ      BÆ    ò      
Ç    ó      ÒÇ    ô      šÈ    õ      bÉ    ö      *Ê    ÷      òÊ    ø      ºË    ù      ‚Ì    ú      JÍ    û      Î    ü      ÚÎ    ý      ¢Ï    þ      jÐ    ÿ      2Ñ           úÑ          ÒÚ          ªã          ‚ì          Zõ          "ö          êö          ²÷          zø    	      Bù    
      
ú          Òú          šû          bü          *ý          òý          ºþ          ‚ÿ          J                     Ú          ¢          j          2          ú          Â          Š          R                    â          ª	          r
           :    !          "      Ê    #      ’    $      Z    %      "    &      ê    '      ²    (      z    )      B    *      
    +      Ò    ,      š    -      b    .      *    /      ò    0      º    1      ‚    2      J    3          4      Ú    5      ¢    6      j    7      2    8      ú    9      Â    :      Š    ;      R     <      !    =      â!    >      ª"    ?      r#    @      :$    A      -    B      ê5    C      Â>    D      šG    E      bH    F      *I    G      òI    H      ºJ    I      ‚K    J      JL    K      M    L      ÚM    M      ¢N    N      jO    O      2P    P      úP    Q      ÂQ    R      ŠR    S      RS    T      T    U      âT    V      ªU    W      rV    X      :W    Y      X    Z      ÊX    [      ’Y    \      ZZ    ]      "[    ^      ê[    _      ²\    `      z]    a      B^    b      
_    c      Ò_    d      š`    e      ba    f      *b    g      òb    h      ºc    i      ‚d    j      Je    k      f    l      Úf    m      ¢g    n      jh    o      2i    p      úi    q      Âj    r      Šk    s      Rl    t      m    u      âm    v      ªn    w      ro    x      :p    y      q    z      Êq    {      ’r    |      Zs    }      "t    ~      êt          ²u    €      zv    �      R    ‚      *ˆ    ƒ      ‘    „      Ú™    …      ¢š    †      j›    ‡      2œ    ˆ      úœ    ‰      Â�    Š      Šž    ‹      RŸ    Œ           �      â     Ž      ª¡    �      r¢    �      :£    ‘      ¤    ’      Ê¤    “      ’¥    ”      Z¦    •      "§    –      ê§    —      ²¨    ˜      z©    ™      Bª    š      
«    ›      Ò«    œ      š¬    �      b­    ž      *®    Ÿ      ò®           º¯    ¡      ‚°    ¢      J±    £      ²    ¤      Ú²    ¥      ¢³    ¦      j´    §      2µ    ¨      úµ    ©      Â¶    ª      Š·    «      R¸    ¬      ¹    ­      â¹    ®      ªº    ¯      r»    °      :¼    ±      ½    ²      Ê½    ³      ’¾    ´      Z¿    µ      "À    ¶      êÀ    ·      ²Á    ¸      zÂ    ¹      BÃ    º      
Ä    »      ÒÄ    ¼      šÅ    ½      bÆ    ¾      *Ç    ¿      òÇ       ç   8        5   U          9   I          I   Y         U   u          Y   i          q   �            Ÿ          �   ‘          –   ¦         Ÿ   ¿          ¦   ¶         ¿   ß         à                         z   reduce_sum_f16     Ç   0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �   "   �   š   ‰   "   ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  64      input0Elements  64      input1Elements  448     usefulCycles    112     physicalWorkCycles      112     workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 64]                         €       :         b      J      7   Q     operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1               5   =         =   E         E   M         M   U         U   ]         ]   e         e   m         m   u         u   }         }   …         …   �         �   •         •   �         �   ¥         ¥   ­         ­   µ         µ   ½         ½   Å         Å   Í         Í   Õ         Õ   Ý         Ý   å         å   í         í   õ         õ   ý         ý                                           %        %  -        -  5        5  =        =  E        E  M        M  U        U  ]        ]  e        e  m        m  u        u  }        }  …        …  �        �  •        •  �        �  ¥        ¥  ­        ­  µ        µ  ½        ½  Å        Å  Í        Í  Õ        Õ  Ý        Ý  å        å  í        í  õ        õ  ý        ý                                          %        %  -        -  5                                   
      Š      g   ±              copy_strided_u64              -   :   -   R   1   2   1   "   1   B   1   z  U   b   Y      Y   š   a   *   a   R   e   ‚   reason  LocalCopy       bytes   128     pattern Strided { rows: 4, row_bytes: 32, source_stride: 32, destination_stride: 128 }  invocations     1       uniformInvocations      true    arguments       [4, 4, 32, 128]                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ñŸ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡¯Õ”    ¯ÕCØ„    CØ©Ûü    ©Û•Ý�    •ÝñÞÐ    ñÞ!âˆ    !âã°    ã›éÜ    ›é/ìô	    /ìWîl    Wî£ð@    £ð§ò€    §òû1                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         %   ¥         5   µ         ¼   Ä         Ì   Ô         í   õ                   -  5        M  U        m  u        �  •        ­  µ        Í  Õ        í  õ                  -  5        M  U        m  u        �  •        ­  µ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1               5   e          6   V          V   v         e   •          v   –         •   Å                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   §   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       (         !   )          )   1          3   ;         5   =          ;   C          C   I          I   Q         N   T          V   X         j   n                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        A   Á          ¬   ´         Ç   Ï         Ï   ×         ×   ß         ì   ô         õ   û                                                    (  0        =  ?        ?  A        A  I        K  S         X  \        \  d        ”  š        š  ¢        ¢  ª        ª  ²        ²  º        º  Â      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8        5   U          8   H          H   X         U   u          Y   i          p   €         ~   ž          ‚   ’          ”   ¤         ž   ¾          ¤   ´         ¾   Þ         Þ   þ         þ                                b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �   "   �   š   ‰   "   ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  64      input0Elements  64      input1Elements  448     usefulCycles    112     physicalWorkCycles      112     workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 64]             €       :         b      J      7   Q   G   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                   %          5   =          M   U          %  -                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ñŸ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡¯Õ”    ¯ÕCØ„    CØ©Ûü    ©Û•Ý�    •ÝëÞÐ    ëÞâ�    âã¸    ã›éä    ›é/ìü	    /ìWît    Wî£ðH    £ð­òˆ    ­ò2                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         +   «         7   ·         »   Ã         Ë   Ó         ì   ô                   ,  4        L  T        l  t        Œ  ”        ¬  ´        Ì  Ô        ì  ô                  ,  4        L  T        l  t        Œ  ”        ¬  ´                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                5   U         7   g          U   u         g   —          u   •         —   Ç                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   ·   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       ,                      #   +          -   3          3   9         7   ?          9   A          A   I          I   Q          Q   S         S   W         a   g                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        F   Æ         Æ   Î         Î   Ö         Ö   Þ          ã   ë         ë   ó         ô   ú                                       '  /         4  8        <  >        >  @        @  H        J  R        [  c         h  n        “  ™        ™  ¡        £  «        «  ³        ³  »        »  Ã      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8        7   W          9   I          I   Y         W   w          Y   i          q   �         �   ¡          ‚   ’          ’   ¢         ¡   Á          §   ·         Á   á         á             !                             b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �   "   �   š   ‰   "   ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  64      input0Elements  64      input1Elements  448     usefulCycles    112     physicalWorkCycles      112     workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 64]             €       :         b      J      7   Q   G   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                #   +          ;   C          S   [          +  3                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ñŸ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡¯Õ”    ¯ÕCØ„    CØ©Ûü    ©Û•Ý�    •ÝëÞÐ    ëÞÙäˆ    ÙäÏå°    Ïå›éÜ    ›é/ìô	    /ìWîl    Wî£ð@    £ð¹ò€    ¹ò2                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         3   ³         7   ·         ¼   Ä         Ì   Ô         í   õ                   -  5        M  U        m  u        �  •        ­  µ        Í  Õ        í  õ                  -  5        M  U        m  u        �  •        ­  µ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                4   T         7   g          T   t         g   —          t   ”         —   Ç                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   §   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       (            "          "   (          (   0          0   8         7   ?          8   @          @   H          H   J         L   T         `   b                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  20      input0Elements  20      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [20]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        =   ½          ¼   Ä         Ç   Ï         Ï   ×         ×   ß         ì   ô         õ   û                                                              (  0        =  ?        ?  A        A  I        K  S        \  d        ”  š        š  ¢        ¢  ª        ª  ²        ²  º        º  Â      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8        7   W          8   H          H   X         W   w          X   h          p   €         €              �   ‘          ’   ¢             À          ¥   µ         À   à         à                                             b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �   "   �   š   ‰   "   ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  64      input0Elements  64      input1Elements  448     usefulCycles    112     physicalWorkCycles      112     workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 64]             €       :         b      J      7   Q   G   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                +   3          C   K          [   c          3  ;                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ñŸ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                 ýŸ±0²                                      b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                                             b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡¯Õ”    ¯ÕCØ„    CØ¯Ûü    ¯Û›Ý�    ›ÝëÞÐ    ëÞâ�    âã¸    ãwéä    wéìü	    ì÷ít    ÷í­ïH    ­ï¿òˆ    ¿ò2                              b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ˜         9   ¹         »   Ã         Ë   Ó         ì   ô                   ,  4        L  T        l  t        Œ  ”        ¬  ´        Ì  Ô        ì  ô                  ,  4        L  T        l  t        Œ  ”        ¬  ´                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                5   U         9   i          U   u         i   ™          u   •         ™   É                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   ·   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       ,         %   -          -   5          7   ;         9   A          ;   C          C   K          K   M          M   Q         S   [          U   [         b   d                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        9   ¹          ¨   °         Æ   Î         Î   Ö          Ô   Ü         Û   ã         ð   ø                              "  $        3  9        9  A        K  Q        Q  S        S  U        U  ]        ]  e        g  o        o  w        w            ‡        ‡  �        �  —        —  Ÿ      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            9         9   I          E   U         M   ]          U   e         a   q         q   �          r   ‚          ‚   ’         ‰   ™          ’   ¢         ™   ©          ¢   ²         ¬   ¼                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                a   i          9  A                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ÷   <      ¡p     ¡¯Õœ    ¯ÕCØŒ    CØ¯Û    ¯Û›Ý˜    ›ÝåÞØ    åÞ±ßˆ    ±ßáâ<    áâ×ãd    ×ã}é�	    }éì¨
    ìñí     ñí§ïô    §ïÅò4    Åò2¬                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L                       9   ¹         ¼   Ä         Ì   Ô         í   õ                   -  5        M  U        m  u        �  •        ­  µ        Í  Õ        í  õ                  -  5        M  U        m  u        �  •        ­  µ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                4   T         9   i          T   t         i   ™          t   ”         ™   É                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   —   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       $            #          #   +          +   3          3   9          9   =         =   C          >   F          F   H         X   `                              b      ‚      w   �      operation.1     static_copy_u32       5   :   5   R   9   2   9      9   2   9      9   B   9   Z   =   b   A      A   š   I   *   I   R   M   "   reason  LocalCopy       value   16      bytes   16      pattern Contiguous      invocations     1       uniformInvocations      true    arguments       [4]                                   b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]         	     €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        A   Á         Ç   Ï         Ï   ×         Ü   ä          é   ñ         ñ   ù          þ                                #  '        4  :        :  B        L  R        R  T        T  V        V  ^        ^  f        f  n        n  v        v  ~        ~  †        ˆ  �        �  ˜        ˜         
   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            8         9   I          D   T         M   ]          T   d         b   r          p   €         r   ‚          €   �         ‰   ™          �             š   ª              °         ª   º                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                i   q          A  I                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ÷   <      ¡p     ¡¯Õœ    ¯ÕCØŒ    CØ¯Û    ¯Û›Ý˜    ›ÝßÞØ    ßÞà�    à5ãD    5ã+äl    +ä}é˜	    }éì°
    ì÷í(    ÷í­ïü    ­ïËò<    Ëò2´                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         &   ¦         ;   »         »   Ã         Ë   Ó         ì   ô                   ,  4        L  T        l  t        Œ  ”        ¬  ´        Ì  Ô        ì  ô                  ,  4        L  T        l  t        Œ  ”        ¬  ´                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                5   U         ;   k          U   u         k   ›          u   •         ›   Ë                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   §   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       (         "   *          +   3          3   7          7   =          =   A          A   C          C   I          I   K         O   U         U   Y                              b      ‚      w   �      operation.1     static_copy_u32       5   :   5   R   9   2   9      9   2   9      9   B   9   Z   =   b   A      A   š   I   *   I   R   M   "   reason  LocalCopy       value   16      bytes   32      pattern Contiguous      invocations     1       uniformInvocations      true    arguments       [8]                                   b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]         	     €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        F   Æ         Æ   Î         Î   Ö         Û   ã          è   ð         ð   ø          ý                                          3  9        9  A        K  Q        Q  S        S  U        U  ]        ]  e        g  o        o  w        w            ‡        ‡  �        �  —        —  Ÿ      
   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            9         ;   K          E   U         O   _          U   e         c   s          q   �         s   ƒ          �   ‘         ‹   ›          ‘   ¡         �   ­          ¡   ±         ­   ½                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                o   w          G  O                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ÷   <      ¡p     ¡µÕœ    µÕIØŒ    IØ¯Û    ¯Û›Ý˜    ›ÝñÞØ    ñÞ�ß�    �ß—äD    —ä�ål    �å}é˜	    }éì°
    ì÷í(    ÷í­ïü    ­ï×ò<    ×ò%2´                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         .   ®         ;   »         ¼   Ä         Ì   Ô         í   õ                   -  5        M  U        m  u        �  •        ­  µ        Í  Õ        í  õ                  -  5        M  U        m  u        �  •        ­  µ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                4   T         ;   k          T   t         k   ›          t   ”         ›   Ë                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  96      input0Elements  96      input1Elements  288     usefulCycles    72      physicalWorkCycles      72      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 96]             €       ˆ          b      J      7   Q   §   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       (                       !   )          )   1          1   7          7   9          9   =          =   C          C   G         L   T         ^   f                              b      ‚      w   �      operation.1     static_copy_u32       5   :   5   R   9   2   9      9   2   9      9   B   9   Z   =   b   A      A   š   I   *   I   R   M   "   reason  LocalCopy       value   16      bytes   8       pattern Contiguous      invocations     1       uniformInvocations      true    arguments       [2]                                   b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   "   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  4       input0Elements  4       usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [4]               	                    b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]         	     €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        =   ½          ¨   °         Ç   Ï         Ï   ×         Ü   ä         ñ   ù          õ   ý          	                              4  :        :  B        L  R        R  T        T  V        V  ^        ^  f        f  n        n  v        v  ~        ~  †        ˆ  �        �  ˜        ˜         
                       b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            8         ;   K          D   T         O   _          T   d         c   s          p   €         s   ƒ          €   �         ‹   ›          �             œ   ¬              °         ¬   ¼                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                w             O  W                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÁÕ”    ÁÕUØ„    UØ—Ûü    —ÛƒÝ�    ƒÝñÞÐ    ñÞ!âx    !âã     ã•éÌ    •é)ìä	    )ì	î\    	î¿ï0    ¿ïÝòp    Ýò+2è                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            —         =   ½         Ã   Ë         Ü   ä         ü             $        <  D        \  d        |  „        œ  ¤        ¼  Ä        Ü  ä        ü            $        <  D        \  d        |  „        œ  ¤        ¼  Ä                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   E         =   ]          N   n         e   …          n   Ž         Ž   ®                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   ‡   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                 (   0          5   =         =   E          >   D          D   L         J   R          L   N         g   i                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        =   ½          À   È         Ä   Ì         Ì   Ô         Ô   Ø         Ø   à         à   è         è   ì         ì   ô                                                  !  )         &  .        )  1        1  9        B  J        X  `        d  l        w            ‡        ˆ  �         »  ½      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          1   A         =   M          I   Y         Q   a          [   k         i   y          l   |         y   ‰          ~   Ž         ‰   ™         š   ª          Ÿ   ¿         ¾   Î                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                }   …          U  ]                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÁÕ”    ÁÕUØ„    UØ—Ûü    —ÛƒÝ�    ƒÝßÐ    ß3â€    3â)ã¨    )ãkéÔ    kéÿë(	    ÿë÷íd	    ÷í­ïJ     ­ïãòâ     ãò72ê                              b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            Ÿ         =   ½         Ä   Ì         Ý   å         ý             %        =  E        ]  e        }  …        �  ¥        ½  Å        Ý  å        ý            %        =  E        ]  e        }  …        �  ¥        ½  Å                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   D         =   ]          N   n         e   …          n   Ž         �   ¯                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   —   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       $         ,   4          4   8          8   <          =   ?         @   H          A   G          G   O         R   V         r   x                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '         operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       	   	                    b      ª  ’     %      operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18           
     €       &         b      J      7   Z     operation.2     exchange                 :      z        "     2     J     reason  OperatorInputs     ‡  `        C   Ã         Å   Í         Í   Õ         Õ   Ù         Ù   á         á   é         é   í         í   õ                                         !        "  *        *  2        2  :         ?  C        C  K         P  X        Y  a        c  k         p  v        v  ~        ~  †        ‡  �         Ç   0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []         2   value         20         b   invocations           1          ç   8            (          0   @         =   M          H   X         Q   a          Y   i         i   y          k   {         y   ‰          |   Œ         ‰   ™         ™   ©          �   ½         ­   ½                                    b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]                   €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                …   �          ]  e                                     b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÁÕ”    ÁÕUØ„    UØ—Ûü    —ÛƒÝ�    ƒÝëÞÐ    ëÞóåx    óåéæ     éæ‰éÌ    ‰éìä	    ì	î\    	î¿ï0    ¿ïéòp    éò=2è                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         %   ¥         ?   ¿         Ã   Ë         Ü   ä         ü             $        <  D        \  d        |  „        œ  ¤        ¼  Ä        Ü  ä        ü            $        <  D        \  d        |  „        œ  ¤        ¼  Ä                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   E         ?   _          M   m         g   ‡          m   �         �   ¯                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   ‡   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                 )   1          1   9          =   E         ?   G          E   I          J   N         S   Y         [   _                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  24      input0Elements  24      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [24]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        D   Ä         Ä   Ì         Ì   Ô         Ô   Ø         Ø   à         à   è         è   ì         ì   ô                                                  !  )        )  1        1  9        B  J         P  X        X  `        d  l         q  w        w            ‡        ˆ  �         ª  ®      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          1   A         ?   O          I   Y         S   c          [   k         k   {          l   |         {   ‹             Ÿ         ‹   ›         œ   ¬          Ÿ   ¯         ¿   Ï                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ‹   “          c  k                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÇÕ”    ÇÕ[Ø„    [Ø—Ûü    —ÛƒÝ�    ƒÝ÷ÞÐ    ÷Þ'â€    'âã¨    ãkéÔ    kéÿëì	    ÿëýíd    ýí³ï8    ³ïõòx    õòC2ð                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         -   ­         ?   ¿         Ä   Ì         Ý   å         ý             %        =  E        ]  e        }  …        �  ¥        ½  Å        Ý  å        ý            %        =  E        ]  e        }  …        �  ¥        ½  Å                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   D         ?   _          L   l         g   ‡          l   Œ         �   ¯                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   —   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1       $         -   5          5   =          =   A         B   J          C   I          I   M          M   O         \   d         m   o                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        ?   ¿         Å   Í          Ê   Ò         Í   Õ         Õ   Ù         Ù   á         á   é         é   í         í   õ                                                    !        "  *        *  2        2  :         3  ;        C  K        Y  a        c  k        v  ~        ~  †        ‡  �      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          0   @         ?   O          H   X         S   c          Y   i          i   y         k   {         {   ‹          }   �         ‹   ›         ›   «          ¬   ¼         ¯   ¿                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                “   ›          k  s                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ¥0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÇÕ”    ÇÕ[Ø„    [Ø�Ûü    �Û‰Ý�    ‰ÝñÞÐ    ñÞ!âx    !âã     ãwéÌ    wéìä	    ìî\    îÅï0    Åïûòp    ûòI2è                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ’         A   Á         Ã   Ë         Ü   ä         ü             $        <  D        \  d        |  „        œ  ¤        ¼  Ä        Ü  ä        ü            $        <  D        \  d        |  „        œ  ¤        ¼  Ä                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   E         A   a          M   m         i   ‰          m   �         ‘   ±                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   ‡   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                 &   .          /   7          7   =          =   E         A   I          E   G         R   Z         b   d                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        A   Á          °   ¸         Ä   Ì         Ì   Ô         Ô   Ü         Ü   ä         ä   ê         ê   î         ò   ô         ô   ü            
                    2  4        7  ;        >  @        F  N        P  X        X  `        a  i        l  t        t  |        |  „        „  Œ        Ž  –      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9         A   Q          I   Y         Q   a          [   {         m   }          {   ‹         ~   Ž         ™   ©          š   ª          ª   º         ®   ¾         ¿   Ï                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ™   ¡          q  y                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÇÕ”    ÇÕ[Ø„    [Ø�Ûü    �Û‰Ý�    ‰ÝåÞÐ    åÞâx    âã     ãqéÌ    qéìä	    ì	î\    	î¿ï0    ¿ïóp    óU2è                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            š         A   Á         Ä   Ì         Ý   å         ý             %        =  E        ]  e        }  …        �  ¥        ½  Å        Ý  å        ý            %        =  E        ]  e        }  …        �  ¥        ½  Å                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   D         A   a          L   l         i   ‰          l   Œ         ‘   ±                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   ‡   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                 *   2          3   ;          ;   C         A   I          C   I          N   P         P   V         X   \                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        C   Ã         Å   Í         Í   Õ         Õ   Ý         Ý   å         å   ë         ë   ï         ó   õ         õ   ý            
                     !         *  0        8  <        ?  A        G  O        O  W        W  _        `  h        m  u        u  }        }  …        …  �        �  •      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8         A   Q          H   X         Q   a          Y   y         m   }          y   ‰            �          ˜   ¨         ›   «          ¨   ¸         ­   ½         ½   Í                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ¡   ©          y  �                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÇÕ”    ÇÕ[Ø„    [Ø�Ûü    �Û‰Ý�    ‰ÝñÞÐ    ñÞóåp    óåéæ˜    éæwéÄ    wéìÜ	    ì	îT    	î¿ï(    ¿ïóh    ó[2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L                       C   Ã         Ã   Ë         Ü   ä         ü             $        <  D        \  d        |  „        œ  ¤        ¼  Ä        Ü  ä        ü            $        <  D        \  d        |  „        œ  ¤        ¼  Ä                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   E         C   c          M   m         k   ‹          m   �         “   ³                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                '   /          /   7          7   ?          ?   G         C   K         [   a         a   e                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   "   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  8       input0Elements  8       usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [8]                                   b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        D   Ä         Ä   Ì         Ì   Ô         Ô   Ü         Ü   ä         ä   ê         ê   î         ò   ô         ô   ü            	                               %  +        7  ;        >  @        F  N        P  X        X  `        a  i        l  t        t  |        |  „        „  Œ        Ž  –      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9         C   S          I   Y         S   c          [   {         o             {   ‹            �          ‹   ›         �   Ÿ          ›   «         ¬   ¼         ¼   Ì                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                §   ¯            ‡                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8      ¡h     ¡ÇÕ”    ÇÕ[Ø„    [Ø�Ûü    �Û‰Ý�    ‰ÝýÞÐ    ýÞ-âp    -â#ã˜    #ãwéÄ    wéìÜ	    ì	îT    	î¿ï(    ¿ïóh    óa2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         (   ¨         C   Ã         Ä   Ì         Ý   å         ý             %        =  E        ]  e        }  …        �  ¥        ½  Å        Ý  å        ý            %        =  E        ]  e        }  …        �  ¥        ½  Å                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   D         C   c          L   l         k   ‹          l   Œ         “   ³                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                +   3          3   ;          ;   C         C   K          D   L         d   l         o   q                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        C   Ã          ¶   ¾         Å   Í         Í   Õ         Õ   Ý         Ý   å         å   ë         ë   ï         ó   õ         õ   ý          û             !         -  /        8  <        ?  A        G  O        O  W        W  _        `  h        m  u        u  }        }  …        …  �        �  •      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8         C   S          H   X         S   c          Y   y         o             y   ‰            �          ‰   ™         �   Ÿ          ™   ©         ­   ½         ½   Í                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ¯   ·          ‡  �                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÷Ÿ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ—Ûü    —ÛƒÝ�    ƒÝ÷ÞÐ    ÷Þ'âp    'âã˜    ã¿éÄ    ¿éSìÜ	    SìîT    îÑï(    Ñïóh    óg2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ‘         E   Å         Ô   Ü         ô   ü                   4  <        T  \        t  |        ”  œ        ´  ¼        Ô  Ü        ô  ü                  4  <        T  \        t  |        ”  œ        ´  ¼        Ì  Ô                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          =   m         E   e         n   Ž          o   �         Ž   ®                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                1   9          9   A          A   I         G   O          I   Q         _   g         g   i                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        E   Å          Ÿ   §         Ç   Ï         Ï   ×         ×   ß          î   ð                               $           &        $  &        &  .        N  R        p  t        �  –        ˜              ¨        ¨  °        °  ¸        ¸  À        Â  Ê        Ê  Ò        Ò  Ú        Û  Ý      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9          9   I         E   U          U   e         Y   i          e   u         i   y          w   —            �          —   §         ž   ®         ®   ¾         Î   Þ                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                µ   ½          �  •                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ�Ûü    �Û‰Ý�    ‰Ý÷ÞÐ    ÷Þÿåp    ÿåõæ˜    õæ¿éÄ    ¿éSìÜ	    SìîT    îÅï(    Åïóh    ós2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ™         E   Å         Õ   Ý         õ   ý                   5  =        U  ]        u  }        •  �        µ  ½        Õ  Ý        õ  ý                  5  =        U  ]        u  }        •  �        µ  ½        Í  Õ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          <   l         E   e          l   Œ         o   �         �   ¯                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                5   =          =   E          E   M         H   P          M   U         V   \         f   j                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  28      input0Elements  28      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [28]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        E   Å          ½   Å         È   Ð         Ð   Ø         Ø   à          å   é          ð   ö                               %        %  '        '  /        O  S        q  u        ‘  —        —  Ÿ        Ÿ  §        ©  ±        ±  ¹        ¹  Á        Á  É        É  Ñ        Ñ  Ù        Ü  Þ      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8          9   I         E   U          T   d         Y   i          d   t         i   y          u   •            �          •   ¥         �   ­         ­   ½         ½   Í                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ½   Å          •  �                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ�Ûü    �Û‰Ý�    ‰Ý÷ÞÐ    ÷Þ'âp    'âã˜    ã¿éÄ    ¿éSìÜ	    SìîT    îÅï(    Åï%óh    %óy2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            Ÿ         G   Ç         Ô   Ü         ô   ü                   4  <        T  \        t  |        ”  œ        ´  ¼        Ô  Ü        ô  ü                  4  <        T  \        t  |        ”  œ        ´  ¼        Ì  Ô                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          =   m         G   g          m   �         o   �         �   ¯                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                .   6          6   >          >   F         G   M          H   P         Y   ]         a   i                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        G   Ç          ¤   ¬         Ç   Ï         Ï   ×         ×   ß          ä   ê          ñ   õ                               $        $  &        &  .        N  R        p  t        �  –        ˜              ¨        ¨  °        °  ¸        ¸  À        Â  Ê        Ê  Ò        Ò  Ú        Û  Ý      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9          9   I         G   W          U   e         [   k          e   u         k   {         �   ‘          ‡   —          —   ·         Ÿ   ¯         ±   Á         Á   Ñ                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                Ã   Ë          ›  £                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ�Ûü    �Û‰Ý�    ‰ÝýÞÐ    ýÞ-âp    -â#ã˜    #ã¿éÄ    ¿éSìÜ	    SìîT    îÅï(    Åï1óh    1ó2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         '   §         G   Ç         Õ   Ý         õ   ý                   5  =        U  ]        u  }        •  �        µ  ½        Õ  Ý        õ  ý                  5  =        U  ]        u  }        •  �        µ  ½        Í  Õ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          <   l         G   g          l   Œ         o   �         �   ¯                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                2   :          :   B          B   J         J   R          K   S         [   c         l   n                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        G   Ç          ­   µ         È   Ð          Ê   Ò         Ð   Ø         Ø   à          ë   í                               %        %  '        '  /        O  S        q  u        ‘  —        —  Ÿ        Ÿ  §        ©  ±        ±  ¹        ¹  Á        Á  É        É  Ñ        Ñ  Ù        Ü  Þ      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8          8   H         G   W          T   d         [   k          d   t         k   {         €   �          …   •          •   µ         Ÿ   ¯         °   À         À   Ð                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                Ë   Ó          £  «                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ�Ûü    �Û‰Ý�    ‰ÝñÞÐ    ñÞ!âp    !âã˜    ã›éÄ    ›é/ìÜ	    /ìîT    îÅï(    Åï7óh    7ó…2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            Œ         I   É         Ô   Ü         ô   ü                   4  <        T  \        t  |        ”  œ        ´  ¼        Ô  Ü        ô  ü                  4  <        T  \        t  |        ”  œ        ´  ¼        Ì  Ô                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          =   m         I   i          m   �         q   ‘         ‘   ±                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                /   7          7   ?          ?   G          G   O         I   Q         U   W         Z   b                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        I   É          Ÿ   §         Ë   Ó         Ó   Û          â   ê          ù   û         ý                               0  6        ?  C        F  N        N  T        T  V        V  X        p  x        x  €        €  ˆ        ˆ  �        ’  š        š  ¢        ¢  ª        ª  ²        ²  º      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   I         I   Y          M   ]         ]   m          ]   m         m   }          q   �         }   �          ƒ   “         Ž   ž          “   £         ž   ®         ¾   Î                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                Ñ   Ù          ©  ±                               b         "      	      operation.3               Š   repeat-remainder           7            :      :      2            Z         reason  Repeat  value   23      iterations      2                  ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ�Ûü    �Û‰Ý�    ‰ÝñÞÐ    ñÞóåp    óåéæ˜    éæ›éÄ    ›é/ìÜ	    /ìîT    îÅï(    Åï=óh    =ó‘2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ”         I   É         Õ   Ý         õ   ý                   5  =        U  ]        u  }        •  �        µ  ½        Õ  Ý        õ  ý                  5  =        U  ]        u  }        •  �        µ  ½        Í  Õ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          <   l         I   i          l   Œ         q   ‘         ‘   ±                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                3   ;          ;   C          C   K         I   Q          K   S         U   Y         \   b                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  12      input0Elements  12      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [12]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        I   É          ¿   Ç         Ì   Ô         Ô   Ü          á   ç          ç   ë         þ           	                     1  7        @  D        G  O        O  U        U  W        W  Y        q  y        y  �        �  ‰        ‰  ‘        ‘  ™        ™  ¡        ¡  ©        ©  ±        ±  ¹      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   H         I   Y          L   \          \   l         ]   m         m   }          p   €         }   �          €   �         �   �         �   ­          ¡   ±         ½   Í                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                Ù   á          ±  ¹                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ£Ûü    £Û�Ý�    �ÝýÞÐ    ýÞ-âp    -â#ã˜    #ã›éÄ    ›é/ìÜ	    /ìîT    îËï(    ËïCóh    Có—2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            š         K   Ë         Ô   Ü         ô   ü                   4  <        T  \        t  |        ”  œ        ´  ¼        Ô  Ü        ô  ü                  4  <        T  \        t  |        ”  œ        ´  ¼        Ì  Ô                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          =   m         K   k          m   �         s   “         “   ³                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                ,   4          4   <          <   D          D   L         K   S         X   \         e   k                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        K   Ë          ª   ²          ¶   ¼         Ë   Ó         Ó   Û          é   í         ý                               0  6        ?  C        F  N        N  T        T  V        V  X        p  x        x  €        €  ˆ        ˆ  �        ’  š        š  ¢        ¢  ª        ª  ²        ²  º      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   I         K   [          M   ]          ]   m         _   o         o             q   �            �          Ž   ž         œ   ¬          Ÿ   ¯         ¯   ¿         Á   Ñ                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ß   ç          ·  ¿                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡ÙÕ”    ÙÕmØ„    mØ£Ûü    £Û�Ý�    �ÝßÐ    ß3âp    3â)ã˜    )ã›éÄ    ›é/ìÜ	    /ì!îT    !î×ï(    ×ïOóh    Oó�2à                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         "   ¢         K   Ë         Õ   Ý         õ   ý                   5  =        U  ]        u  }        •  �        µ  ½        Õ  Ý        õ  ý                  5  =        U  ]        u  }        •  �        µ  ½        Í  Õ                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          <   l         K   k          l   Œ         s   “         “   ³                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  64      input0Elements  64      input1Elements  192     usefulCycles    48      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   w   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                0   8          8   @          @   H          H   P         K   S         `   h         p   r                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        K   Ë          ¥   ­          Ä   Æ         Ì   Ô         Ô   Ü          î   ö         þ           	                     1  7        @  D        G  O        O  U        U  W        W  Y        q  y        y  �        �  ‰        ‰  ‘        ‘  ™        ™  ¡        ¡  ©        ©  ±        ±  ¹      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   H         K   [          L   \          \   l         _   o         o             p   €            �          Ž   ž         œ   ¬          ž   ®         ¿   Ï         Ï   ß                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ç   ï          ¿  Ç                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ«0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡åÕ”    åÕyØ„    yØ©Ûü    ©Û•Ý�    •ÝýÞÐ    ýÞäh    äýä�    ýä³é¼    ³éGìÔ	    GìîL    îËï     ËïUó`    Uó£2Ø                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ‹         M   Í         ä   ì                   $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        ä  ì                  $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        Ô  Ü                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          5   U         N   n          e   •         n   Ž         —   ·                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                '   /          /   7          7   ?         M   U         a   i         l   n                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [32]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        M   Í          Ÿ   §          ½   Å         Ï   ×         ×   ß                                                              &        &  .        4  6        _  c        ‚  ˆ        ˆ  �        ’  š        š  ¢        ¢  ª        ª  ²        ´  ¼        ¼  Ä        Ä  Ì        Ì  Ð      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9          9   I         M   ]          N   ^         a   q          i   y         q   �          }   �         Œ   œ         Ÿ   ¯         ¯   ¿          °   Ð         ¿   Ï                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                í   õ          Å  Í                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡åÕ”    åÕyØ„    yØ£Ûü    £Û�Ý�    �ÝßÐ    ß3âh    3â)ã�    )ã³é¼    ³éGìÔ	    Gì!îL    !î×ï     ×ï[ó`    [ó¯2Ø                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            “         M   Í         å   í                   %  -        E  M        e  m        …  �        ¥  ­        Å  Í        å  í                  %  -        E  M        e  m        …  �        ¥  ­        Å  Í        Õ  Ý                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          4   T         O   o          d   ”         o   �         •   µ                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                #   +          +   3          3   ;         M   U         _   e         m   q                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        M   Í          ¤   ¨          «   ³          Ä   Ê         Ð   Ø         Ø   à                                                   '        '  /        5  7        `  d        ƒ  ‰        ‰  ‘        ‘  ™        ™  ¡        £  «        «  ³        ³  »        »  Ã        Ã  Ë        Í  Ñ      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8          8   H         M   ]          N   ^         a   q          i   y         q   �          |   Œ         Œ   œ         ž   ®          ­   Í         ¾   Î         Î   Þ                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                õ   ý          Í  Õ                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡åÕ”    åÕyØ„    yØ©Ûü    ©Û•Ý�    •ÝýÞÐ    ýÞ-âh    -â#ã�    #ã³é¼    ³éGìÔ	    Gì!îL    !î×ï     ×ïaó`    aóµ2Ø                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ™         O   Ï         ä   ì                   $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        ä  ì                  $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        Ô  Ü                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          5   U         O   o          e   •         o   �         —   ·                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                &   .          .   6          6   >         O   W         [   a         g   k                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        O   Ï          š   ¢          ¶   º         Ï   ×         ×   ß          ä   ê                                                   &        &  .        4  6        _  c        ‚  ˆ        ˆ  �        ’  š        š  ¢        ¢  ª        ª  ²        ´  ¼        ¼  Ä        Ä  Ì        Ì  Ð      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9          9   I          M   ]         O   _         c   s          k   {         s   ƒ          }   �         Ž   ž         ¡   ±          ¯   Ï         ±   Á         Í   Ý                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                û            Ó  Û                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ! ¡h     ¡åÕ”    åÕyØ„    yØ©Ûü    ©Û•Ý�    •ÝýÞÐ    ýÞ-âh    -â#ã�    #ã¹é¼    ¹éMìÔ	    Mì!îL    !î×ï     ×ïmó`    mó»2Ø                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L         !   ¡         O   Ï         å   í                   %  -        E  M        e  m        …  �        ¥  ­        Å  Í        å  í                  %  -        E  M        e  m        …  �        ¥  ­        Å  Í        Õ  Ý                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          4   T         O   o          d   ”         o   �         —   ·                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                "   *          *   2          2   :         O   W         ^   f         j   l                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        O   Ï          ¥   ­         Ð   Ø          Ö   Þ         Ø   à          ò   ô                                                   '        '  /        5  7        `  d        ƒ  ‰        ‰  ‘        ‘  ™        ™  ¡        £  «        «  ³        ³  »        »  Ã        Ã  Ë        Í  Ñ      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8          8   H          L   \         O   _         c   s          i   y         s   ƒ          |   Œ         Œ   œ             °          ­   Í         °   À         Í   Ý                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  32      input0Elements  32      input1Elements  224     usefulCycles    56      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q   '   operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                           Û  ã                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ýŸ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ±0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ' ¡h     ¡åÕ”    åÕyØ„    yØ©Ûü    ©Û•Ý�    •ÝýÞÐ    ýÞëäh    ëäáå�    áå§é¼    §é;ìÔ	    ;ì-îL    -îãï     ãïsó`    sóÁ2Ð                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            †         Q   Ñ         ä   ì                   $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        ä  ì                  $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        Ô  Ü                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          5   U         Q   q          e   •         q   ‘         ™   ¹                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                !   )          )   1          1   9         Q   Y         ]   _         d   l                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   2   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  16      input0Elements  16      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     2       uniformInvocations      false   arguments       [16]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        Q   Ñ          š   ¢          §   ¯         Ó   Û         Û   ã         å   í         í   õ                                %        %  '        '  )        )  1        1  5        7  ;        r  z        z  ‚        ‚  Š        Œ  ”        ”  œ        œ  ¤        ¤  ¬        ®  ¶        ¶  ¾      	   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9          9   I          K   [         Q   a         a   q          f   v          x   ˆ         {   ‹         �   �         �   ­         ­   ½         ¾   Î          Ì   ì                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  16      input0Elements  16      input1Elements  224     usefulCycles    28      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q      operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                á  é                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚßè     ß!â`    !âƒéˆ    ƒé=íø    =í�ñX    �ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               ]   e         i   q                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   ˜          j  r         š  ¢           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ç   8     ' ¡h     ¡åÕ”    åÕyØ„    yØ©Ûü    ©Û•Ý�    •ÝßÐ    ß3âh    3â)ã�    )ã¡é¼    ¡é5ìÔ	    5ì'îL    'îÝï     Ýïyó`    yóÍ2Ð                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            Ž         Q   Ñ         å   í                   %  -        E  M        e  m        …  �        ¥  ­        Å  Í        å  í                  %  -        E  M        e  m        …  �        ¥  ­        Å  Í        Õ  Ý                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          4   T         Q   q          d   ”         q   ‘         ™   ¹                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                   %          %   -          -   5         T   \         b   h         l   p                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        Q   Ñ          ˜              ¶   º         Ô   Ü         Ü   ä         æ   î         î   ö          ÿ                        &        &  (        (  *        *  2        2  6        8  <        s  {        {  ƒ        ƒ  ‹        ‹  “        “  ›        ›  £        ¥  ­        ­  µ        µ  ½      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8          8   H          J   Z         Q   a         a   q          d   t          u   …         z   Š         �   �         �   ­         ­   ½          É   é         Í   Ý                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  16      input0Elements  16      input1Elements  224     usefulCycles    28      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q      operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                é  ñ                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚßè     ß!â`    !âwéˆ    wé=íø    =í�ñX    �ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               ]   e         i   q                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   œ          d  l         Œ  ”           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                          ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     ' ¡h     ¡åÕ”    åÕyØ„    yØ¯Ûü    ¯Û›Ý�    ›ÝßÐ    ß3âh    3â)ã�    )ã§é¼    §é;ìÔ	    ;ì-îL    -îãï     ãïó`    óÓ2Ð                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A   7  operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       L            ”         S   Ó         ä   ì                   $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        ä  ì                  $  ,        D  L        d  l        „  Œ        ¤  ¬        Ä  Ì        Ô  Ü                             b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   5          5   U         S   s          e   •         s   “         ›   »                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                    (          (   0          0   8         S   [         _   c         g   m                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]              €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        S   Ó          ™   ¡         Ó   Û         Û   ã         å   í         í   õ          ú                                   %        %  '        '  )        )  1        1  5        7  ;        r  z        z  ‚        ‚  Š        Œ  ”        ”  œ        œ  ¤        ¤  ¬        ®  ¶        ¶  ¾      	   	                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []             
     €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            )          )   9          9   I          K   [         S   c         c   s          f   v          w   ‡         }   �         �   Ÿ         Ÿ   ¯         ¯   ¿         Á   Ñ          Ë   ë                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  16      input0Elements  16      input1Elements  224     usefulCycles    28      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q      operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ï  ÷                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚ	ßè     	ß'â`    'âkéˆ    ké=íø    =í�ñX    �ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               _   g         k   s                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   Ž          E  M         �  ‰           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ÷   <     ' ¡p     ¡åÕœ    åÕe×|    e×ùÙ0    ùÙ¯Û¨    ¯Û›Ý<    ›Ý	ß|    	ß9â    9â/ã<    /ã§éh	    §é;ì€
    ;ì-îø    -îãïÌ    ãï‹ó    ‹óÙ2|                             b      r      §        operation.0     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   10      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]               €       â         b      J      '   A     operation.0     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       D        å   í                   %  -        E  M        e  m        …  �        ¥  ­        Å  Í        å  í                  %  -        E  M        e  m        …  �        ¥  ­        Å  Í        Õ  Ý                             b      J      w   �      operation.0     copy_u64              5   :   5   R   9   2   9      9   2   9   "   9   B   9   Z   =   b   A      A   š   I   *   I   R   M   B   reason  LocalCopy       value   13      bytes   512     pattern Contiguous      invocations     1       uniformInvocations      true    arguments       [10, 4]                               b      ª  )   Ç   a     operation.0     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   13      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       Ð          b      J      7   M   g   operation.0     exchange                 :      z      2            b         reason  OperatorInputs  value   12      invocations     1                   4          4   T         S   s          d   ”         s   “         ›   »                              b      z      Ç   )     operation.0     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   12      kernelSpec      ReductionSum { partials: 4 }    outputElements  48      input0Elements  48      input1Elements  192     usefulCycles    36      physicalWorkCycles      48      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [3, 64]             €       ˆ          b      J      7   Q   g   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1                   $          $   ,          ,   4         V   ^         k   s         s   u                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i   "   i   š   q   "   q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  36      input0Elements  36      usefulCycles    108     physicalWorkCycles      108     workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [36]                                  b      r      §        operation.2     fill_zero_u64   (      M   :   M   ¢   U   2   U      U   Z   Y   Â  q   z   u   "   u   j   y      y   š   �      �   ’   ‰   r  �   b   ¡      ¡   š   ©   *   ©   R   ­   :   reason  LayoutRearrangement     value   18      kernelSpec      FillZero { offset: 544, bytes: 32, padding_only: true } outputElements  272     usefulCycles    0       physicalWorkCycles      0       workEstimateBasis       padding initialization: no useful tensor work   invocations     1       uniformInvocations      true    arguments       [0, 4]         	     €       ã         b      J      '   A   ‡  operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1       `        S   Ó          �   ¥         Ô   Ü         Ü   ä         æ   î          è   ð         î   ö                                  &        &  (        (  *        *  2        2  6        8  <        s  {        {  ƒ        ƒ  ‹        ‹  “        “  ›        ›  £        ¥  ­        ­  µ        µ  ½      
   
                    b      ª  )   Ç   a     operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18    0      ]   :   ]   z   a   2   a      a   Z   e   ª  �   z   ¡   "   ¡   z   ¥   "   ¥   z   ©   "   ©   j   ­      ­   š   µ      µ   ’   ½   º   Å   b   É      É   š   Ñ   *   Ñ   R   Õ      reason  OperatorKernel  value   21      kernelSpec      Gemm { multiply: F16, accumulate: F16, mode: Initialize, weights: Interleaved, inner_block: 16, output_columns: 16 }    outputElements  272     input0Elements  272     input1Elements  256     usefulCycles    68      physicalWorkCycles      72      workEstimateBasis       AMP FLOPs / IPU21 peak  invocations     1       uniformInvocations      true    arguments       []                  €       &         b      J      7   M   ç   operation.2     exchange                 :      z      2            b         reason  OperatorInputs  value   20      invocations     1       8            (          (   8          8   H          I   Y         S   c         c   s          d   t          u   …         |   Œ         �   Ÿ         Ÿ   ¯         ¯   ¿         À   Ð          É   é                              b      z      Ç   )     operation.2     reduce_sum_f16  0      ]   :   ]   z   a   2   a      a   Z   e   ê   q   z   u      u   z   y      y   z   }   "   }   j   �      �   š   ‰      ‰   ’   ‘   
  ¡   b   ¥      ¥   š   ­   *   ­   R   ±   B   reason  OperatorKernel  value   20      kernelSpec      ReductionSum { partials: 8 }    outputElements  16      input0Elements  16      input1Elements  224     usefulCycles    28      physicalWorkCycles      56      workEstimateBasis       vector reduction add issue slots        invocations     1       uniformInvocations      true    arguments       [7, 32]             €       :         b      J      7   Q      operation.2     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   22      invocations     1                ÷  ÿ                               b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚ	ßè     	ß'â`    'âAéˆ    Aé=íø    =í�ñX    �ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               _   g         k   s                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   ’          *  2         T  \           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                          ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         ' ýÒ8     ýÒéÚˆ     éÚßè     ß!â`    !â;éˆ    ;é=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               U   ]         g   o                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €                     O  W           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚýÞè     ýÞâ`    â_éˆ    _é=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               U   ]         a   i                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   „          d  l         v  ~           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚýÞè     ýÞâ`    âéˆ    é=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               X   `         `   h                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   �          ñ   ù          4  <           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' ýÒ8     ýÒéÚˆ     éÚýÞè     ýÞâ`    âeéˆ    eé=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               U   ]         a   i                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   ˆ          ^  f         z  ‚           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         ' Ó8     ÓéÚˆ     éÚýÞè     ýÞâ`    âYéˆ    Yé=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               W   _         a   i                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   ƒ          >  F         n  v           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß!â`    !â_éˆ    _é=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               W   _         c   k                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          9  A         u  }           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß!â`    !âSéˆ    Sé=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               Z   b         f   n                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   ‡          @  H         h  p           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß!â`    !â5éˆ    5é=íø    =í‡ñX    ‡ñÕ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               W   _         c   k                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          #  +         J  R           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         ' Ó8     ÓéÚˆ     éÚßè     ß-â`    -âAéˆ    Aé=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               Y   a         m   u                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                	   ‰                     S  [           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß!â`    !âwéˆ    wé=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               Y   a         e   m                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   �          Z  b         Š  ’           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß-â`    -â#éˆ    #é=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               Y   a         m   u                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   �          õ   ý          8  @           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß!â`    !âkéˆ    ké=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               Y   a         e   m                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   “          T  \         |  „           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         ' Ó8     ÓéÚˆ     éÚßè     ß-â`    -â_éˆ    _é=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               [   c         o   w                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   �          B  J         r  z           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚ	ßè     	ß'â`    'âYéˆ    Yé=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               [   c         g   o                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   …          -  5         i  q           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚßè     ß-â`    -âYéˆ    Yé=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               [   c         o   w                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   “          D  L         l  t           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         ' Ó8     ÓéÚˆ     éÚ	ßè     	ß'â`    'â)éˆ    )é=íø    =í‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               [   c         g   o                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                	   ‰                     =  E           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ·0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         - Ó8     ÓïÚˆ     ïÚ	ßè     	ß'â`    'âGéˆ    GéCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               ]   e         i   q                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €            #         W  _           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß9â`    9âSéˆ    SéCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               k   s         w                                 b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          T  \         f  n           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚ	ßè     	ß'â`    'â)éˆ    )éCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               ]   e         i   q                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €            
         <  D           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß3â`    3âYéˆ    YéCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               i   q         u   }                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          N  V         j  r           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         - Ó8     ÓïÚˆ     ïÚßè     ß-â`    -âeéˆ    eéCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               _   g         k   s                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          F  N         v  ~           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß3â`    3âMéˆ    MéCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               i   q         u   }                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          !  )         ]  e           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß-â`    -âSéˆ    SéCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               _   g         k   s                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          F  N         b  j           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß3â`    3â#éˆ    #éCíø    Cí‡ñX    ‡ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               g   o         u   }                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €          
           1  9           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        	 ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         - Ó8     ÓïÚˆ     ïÚßè     ß9â`    9âGéˆ    GéCíø    Cí�ñX    �ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               a   i         u   }                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                    €            !         U  ]           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚ!ßè     !ß?â`    ?âkéˆ    kéCíø    Cí�ñX    �ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               g   o         {   ƒ                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   †          J  R         z  ‚           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß9â`    9â/éˆ    /éCíø    Cí�ñX    �ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               a   i         v   ~                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                   „                     @  H           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                 ‡         - Ó8     ÓïÚˆ     ïÚßè     ß9â`    9â_éˆ    _éCíø    Cí�ñX    �ñÛ0¸           €                 b      *      '   9      operation.0     sync             :      ‚      b         reason  ExchangeBarrier invocations     1                   €                 b      *      7   I      operation.0     sync             :      ‚      2            b         reason  ExchangeBarrier value   12      invocations     1                   €       ˆ          b      J      7   Q   '   operation.1     exchange                 :      ¢      2            b   !      reason  LayoutRearrangement     value   15      invocations     1               e   m         y   �                              b      ª      ·        operation.1     gelu_tanh_approx_f16    ,      U   :   U   z   Y   2   Y      Y   Z   ]   *   ]   z   a      a   z   e      e   j   i      i   š   q      q   ’   y   b  �   b   ‘      ‘   š   ™   *   ™   R   �   *   reason  OperatorKernel  value   16      kernelSpec      Gelu    outputElements  32      input0Elements  32      usefulCycles    96      physicalWorkCycles      96      workEstimateBasis       GeLU arithmetic and optional FP8 conversion     invocations     1       uniformInvocations      true    arguments       [32]                €       ã         b      J      '   A   7   operation.2     exchange                 :      ¢      b         reason  LayoutRearrangement     invocations     1                
   Š          D  L         l  t           €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   20      invocations     1                   €                 b      *      7   I      operation.2     sync             :      ‚      2            b         reason  ExchangeBarrier value   22      invocations     1                                       b      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                          ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         ½0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        - Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Û0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ' Õ0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ! Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ï0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         É0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                         Ã0                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23             