   ÿ  ‚  ‚  '  N  `˜                    /hY               D     ï,xå-x€     å-x�1x¬    �1x3x¨    3x•5x\    •5xo7xÔ    o7x[9xh    [9x¥:x¨    ¥:xË;xT	    Ë;xû>x
    û>xñ?x0    ñ?xaExX    aExõGxl    õGxJxä    JxiLxD    iLx{Ox*     {OxûPx     ûPxõ�xº	                              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      '   =   W  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   —   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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      z   ò            operation.2     reduce_sum_f16            \               "              š
            b            )            â)            ª*            r+            :,            -     	       Ê-     
       ’.            Z/            "0            ê0            ²1            z2            B3            
4            Ò4            š5            b6            *7            ò7            º8            ‚9            J:            ;            Ú;            ¢<            j=            2>            ú>             Â?     !       Š@     "       RA     #       B     $       âB     %       ªC     &       rD     '       :E     (       F     )       ÊF     *       ’G     +       ZH     ,       "I     -       êI     .       ²J     /       zK     0       BL     1       
M     2       ÒM     3       šN     4       bO     5       *P     6       òP     7       ºQ     8       ‚R     9       JS     :       T     ;       ÚT     <       ¢U     =       jV     >       2W     ?       úW     @       ÂX     A       Šv     B       Rw     C       
•     D       Ò•     E       š–     F       b—     G       *˜     H       ò˜     I       º™     J       ‚š     K       J›     L             M       Ê      N       ’     O       Z     P       "     Q       ê     R       ²     S       z     T       B     U       
     V       Ò     W       š     X       b	     Y       *
     Z       ò
     [       º     \       ‚     ]       J     ^            _       Ú     `       ¢     a       j     b       2     c       ú     d       Â     e       Š     f       R     g            h       â     i       ª     j       r     k       :     l            m       Ê     n       ’     o       Z     p       "     q       ê     r       ²     s       z     t       B     u       
      v       Ò      w       š!     x       b"     y       *#     z       ò#     {       º$     |       ‚%     }       J&     ~       '            Ú'     €       ¢(     �       JF     ‚       G     ƒ       "f     „       êf     …       ²g     †       zh     ‡       Bi     ˆ       
j     ‰       Òj     Š       šk     ‹       bl     Œ       *m     �       òm     Ž       ºn     �       ‚o     �       Jp     ‘       q     ’       Úq     “       ¢r     ”       js     •       2t     –       út     —       Âu     ˜       Šv     ™       Rw     š       x     ›       âx     œ       ªy     �       rz     ž       :{     Ÿ       |             Ê|     ¡       ’}     ¢       Z~     £       "     ¤       ê     ¥       ²€     ¦       z�     §       B‚     ¨       
ƒ     ©       Òƒ     ª       š„     «       b…     ¬       *†     ­       ò†     ®       º‡     ¯       ‚ˆ     °       J‰     ±       Š     ²       ÚŠ     ³       ¢‹     ´       jŒ     µ       2�     ¶       ú�     ·       ÂŽ     ¸       Š�     ¹       R�     º       ‘     »       â‘     ¼       ª’     ½       r“     ¾       :”     ¿       •     À       Ê•     Á       ê´     Â       ²µ     Ã       ÒÔ     Ä       šÕ     Å       bÖ     Æ       *×     Ç       ò×     È       ºØ     É       ‚Ù     Ê       JÚ     Ë       Û     Ì       ÚÛ     Í       ¢Ü     Î       jÝ     Ï       2Þ     Ð       úÞ     Ñ       Âß     Ò       Šà     Ó       Rá     Ô       â     Õ       ââ     Ö       ªã     ×       rä     Ø       :å     Ù       æ     Ú       Êæ     Û       ’ç     Ü       Zè     Ý       "é     Þ       êé     ß       ²ê     à       zë     á       Bì     â       
í     ã       Òí     ä       šî     å       bï     æ       *ð     ç       òð     è       ºñ     é       ‚ò     ê       Jó     ë       ô     ì       Úô     í       ¢õ     î       jö     ï       2÷     ð       ú÷     ñ       Âø     ò       Šù     ó       Rú     ô       û     õ       âû     ö       ªü     ÷       rý     ø       :þ     ù       ÿ     ú       Êÿ     û       ’     ü       Z    ý       "    þ       ê    ÿ       ²           z          ò!          º"          8          "	           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(     -      *)     .      ò)     /      º*     0      ‚+     1      J,     2      -     3      Ú-     4      ¢.     5      j/     6      20     7      ú0     8      Â1     9      Š2     :      R3     ;      4     <      â4     =      ª5     >      r6     ?      :7     @      8     A      zU     B      BV     C      Ês     D      ’t     E      Zu     F      "v     G      êv     H      ²w     I      zx     J      By     K      
z     L      Òz     M      š{     N      b|     O      *}     P      ò}     Q      º~     R      ‚     S      J€     T      �     U      Ú�     V      ¢‚     W      jƒ     X      2„     Y      ú„     Z      Â…     [      Š†     \      R‡     ]      ˆ     ^      âˆ     _      ª‰     `      rŠ     a      :‹     b      Œ     c      ÊŒ     d      ’�     e      ZŽ     f      "�     g      ê�     h      ²�     i      z‘     j      B’     k      
“     l      Ò“     m      š”     n      b•     o      *–     p      ò–     q      º—     r      ‚˜     s      J™     t      š     u      Úš     v      ¢›     w      jœ     x      2�     y      ú�     z      Âž     {      ŠŸ     |      R      }      ¡     ~      â¡           ª¢     €      r£     �      êÀ     ‚      ²Á     ƒ      *ß     „      òß     …      ºà     †      ‚á     ‡      Jâ     ˆ      ã     ‰      Úã     Š      ¢ä     ‹      jå     Œ      2æ     �      úæ     Ž      Âç     �      Šè     �      Ré     ‘      ê     ’      âê     “      ªë     ”      rì     •      :í     –      î     —      Êî     ˜      ’ï     ™      Zð     š      "ñ     ›      êñ     œ      ²ò     �      zó     ž      Bô     Ÿ      
õ            Òõ     ¡      šö     ¢      b÷     £      *ø     ¤      òø     ¥      ºù     ¦      ‚ú     §      Jû     ¨      ü     ©      Úü     ª      ¢ý     «      jþ     ¬      2ÿ     ­      úÿ     ®      Â     ¯      Š    °      R    ±          ²      â    ³      ª    ´      r    µ      :    ¶          ·      Ê    ¸      ’    ¹      Z	    º      "
    »      ê
    ¼      ²    ½      z    ¾      B    ¿      
    À      Ò    Á      :,    Â      -    Ã      jJ    Ä      2K    Å      úK    Æ      ÂL    Ç      ŠM    È      RN    É      O    Ê      âO    Ë      ªP    Ì      rQ    Í      :R    Î      S    Ï      ÊS    Ð      ’T    Ñ      ZU    Ò      "V    Ó      êV    Ô      ²W    Õ      zX    Ö      BY    ×      
Z    Ø      ÒZ    Ù      š[    Ú      b\    Û      *]    Ü      ò]    Ý      º^    Þ      ‚_    ß      J`    à      a    á      Úa    â      ¢b    ã      jc    ä      2d    å      úd    æ      Âe    ç      Šf    è      Rg    é      h    ê      âh    ë      ªi    ì      rj    í      :k    î      l    ï      Êl    ð      ’m    ñ      Zn    ò      "o    ó      êo    ô      ²p    õ      zq    ö      Br    ÷      
s    ø      Òs    ù      št    ú      bu    û      *v    ü      òv    ý      ºw    þ      ‚x    ÿ      Jy           z          z—          B˜          ªµ          r¶          :·          ¸          Ê¸          ’¹    	      Zº    
      "»          ê»          ²¼          z½          B¾          
¿          Ò¿          šÀ          bÁ          *Â          òÂ          ºÃ          ‚Ä          JÅ          Æ          ÚÆ          ¢Ç          jÈ          2É          úÉ          ÂÊ          ŠË           RÌ    !      Í    "      âÍ    #      ªÎ    $      rÏ    %      :Ð    &      Ñ    '      ÊÑ    (      ’Ò    )      ZÓ    *      "Ô    +      êÔ    ,      ²Õ    -      zÖ    .      B×    /      
Ø    0      ÒØ    1      šÙ    2      bÚ    3      *Û    4      òÛ    5      ºÜ    6      ‚Ý    7      JÞ    8      ß    9      Úß    :      ¢à    ;      já    <      2â    =      úâ    >      Âã    ?      Šä    @      Rå    A      º    B      ‚    C      ê     D      ²!    E      z"    F      B#    G      
$    H      Ò$    I      š%    J      b&    K      *'    L      ò'    M      º(    N      ‚)    O      J*    P      +    Q      Ú+    R      ¢,    S      j-    T      2.    U      ú.    V      Â/    W      Š0    X      R1    Y      2    Z      â2    [      ª3    \      r4    ]      :5    ^      6    _      Ê6    `      ’7    a      Z8    b      "9    c      ê9    d      ²:    e      z;    f      B<    g      
=    h      Ò=    i      š>    j      b?    k      *@    l      ò@    m      ºA    n      ‚B    o      JC    p      D    q      ÚD    r      ¢E    s      jF    t      2G    u      úG    v      ÂH    w      ŠI    x      RJ    y      K    z      âK    {      ªL    |      rM    }      :N    ~      O          ÊO    €      ’P    �      úm    ‚      Ân    ƒ      p    „      :     …           †      Ê     ‡      ’     ˆ      Z      ‰      "!     Š      ê!     ‹      ²"     Œ      z#     �      B$     Ž      
%     �      Ò%     �      š&     ‘      b'     ’      *(     “      ò(     ”      º)     •      ‚*     –      J+     —      ,     ˜      Ú,     ™      ¢-     š      j.     ›      2/     œ      ú/     �      Â0     ž      Š1     Ÿ      R2            3     ¡      â3     ¢      ª4     £      r5     ¤      :6     ¥      7     ¦      Ê7     §      ’8     ¨      Z9     ©      ":     ª      ê:     «      ²;     ¬      z<     ­      B=     ®      
>     ¯      Ò>     °      š?     ±      b@     ²      *A     ³      òA     ´      ºB     µ      ‚C     ¶      JD     ·      E     ¸      ÚE     ¹      ¢F     º      jG     »      2H     ¼      úH     ½      ÂI     ¾      ŠJ     ¿      RK     À      L     Á      ‚i     Â      Jj     Ã      ²‡     Ä      zˆ     Å      B‰     Æ      
Š     Ç      ÒŠ     È      š‹     É      bŒ     Ê      *�     Ë      ò�     Ì      ºŽ     Í      ‚�     Î      J�     Ï      ‘     Ð      Ú‘     Ñ      ¢’     Ò      j“     Ó      2”     Ô      ú”     Õ      Â•     Ö      Š–     ×      R—     Ø      ˜     Ù      â˜     Ú      ª™     Û      rš     Ü      :›     Ý      œ     Þ      Êœ     ß      ’�     à      Zž     á      "Ÿ     â      êŸ     ã      ²      ä      z¡     å      B¢     æ      
£     ç      Ò£     è      š¤     é      b¥     ê      *¦     ë      ò¦     ì      º§     í      ‚¨     î      J©     ï      ª     ð      Úª     ñ      ¢«     ò      j¬     ó      2­     ô      ú­     õ      Â®     ö      Š¯     ÷      R°     ø      ±     ù      â±     ú      ª²     û      r³     ü      :´     ý      µ     þ      Êµ     ÿ      ’¶            Z·           ²Ô           zÕ           Òò           šó           bô           *õ           òõ           ºö     	      ‚÷     
      Jø           ù           Úù           ¢ú           jû           2ü           úü           Âý           Šþ           Rÿ                      â           ª          r          :                    Ê          ’          Z          "          ê          ²           z	    !      B
    "      
    #      Ò    $      š    %      b    &      *    '      ò    (      º    )      ‚    *      J    +          ,      Ú    -      ¢    .      j    /      2    0      ú    1      Â    2      Š    3      R    4          5      â    6      ª    7      r    8      :    9          :      Ê    ;      ’    <      Z    =      "     >      ê     ?      ²!    @      z"    A      Ò?    B      š@    C      ò]    D      º^    E      ‚_    F      J`    G      a    H      Úa    I      ¢b    J      jc    K      2d    L      úd    M      Âe    N      Šf    O      Rg    P      h    Q      âh    R      ªi    S      rj    T      :k    U      l    V      Êl    W      ’m    X      Zn    Y      "o    Z      êo    [      ²p    \      zq    ]      Br    ^      
s    _      Òs    `      št    a      bu    b      *v    c      òv    d      ºw    e      ‚x    f      Jy    g      z    h      Úz    i      ¢{    j      j|    k      2}    l      ú}    m      Â~    n      Š    o      R€    p      �    q      â�    r      ª‚    s      rƒ    t      :„    u      …    v      Ê…    w      ’†    x      Z‡    y      "ˆ    z      êˆ    {      ²‰    |      zŠ    }      B‹    ~      
Œ          ÒŒ    €      š�    �      âª    ‚      ª³    ƒ      òÐ    „      ºÙ    …      ‚Ú    †      JÛ    ‡      Ü    ˆ      ÚÜ    ‰      ¢Ý    Š      jÞ    ‹      2ß    Œ      úß    �      Âà    Ž      Šá    �      Râ    �      ã    ‘      âã    ’      ªä    “      rå    ”      :æ    •      ç    –      Êç    —      ’è    ˜      Zé    ™      "ê    š      êê    ›      ²ë    œ      zì    �      Bí    ž      
î    Ÿ      Òî           šï    ¡      bð    ¢      *ñ    £      òñ    ¤      ºò    ¥      ‚ó    ¦      Jô    §      õ    ¨      Úõ    ©      ¢ö    ª      j÷    «      2ø    ¬      úø    ­      Âù    ®      Šú    ¯      Rû    °      ü    ±      âü    ²      ªý    ³      rþ    ´      :ÿ    µ           ¶      Ê     ·      ’    ¸      Z    ¹      "    º      ê    »      ²    ¼      z    ½      B    ¾      
    ¿      Ò    À      š    Á      â%    Â      ª.    Ã      ZM    Ä      "V    Å      êV    Æ      ²W    Ç      zX    È      BY    É      
Z    Ê      ÒZ    Ë      š[    Ì      b\    Í      *]    Î      ò]    Ï      º^    Ð      ‚_    Ñ      J`    Ò      a    Ó      Úa    Ô      ¢b    Õ      jc    Ö      2d    ×      úd    Ø      Âe    Ù      Šf    Ú      Rg    Û      h    Ü      âh    Ý      ªi    Þ      rj    ß      :k    à      l    á      Êl    â      ’m    ã      Zn    ä      "o    å      êo    æ      ²p    ç      zq    è      Br    é      
s    ê      Òs    ë      št    ì      bu    í      *v    î      òv    ï      ºw    ð      ‚x    ñ      Jy    ò      z    ó      Úz    ô      ¢{    õ      j|    ö      2}    ÷      ú}    ø      Â~    ù      Š    ú      R€    û      �    ü      â�    ý      ª‚    þ      rƒ    ÿ      :„           …          Ê�          ’–          ZŸ          "¨          ê¨          ²©          zª          B«    	      
¬    
      Ò¬          š­          b®          *¯          ò¯          º°          ‚±          J²          ³          Ú³          ¢´          jµ          2¶          ú¶          Â·          Š¸          R¹          º          âº          ª»          r¼          :½           ¾    !      Ê¾    "      ’¿    #      ZÀ    $      "Á    %      êÁ    &      ²Â    '      zÃ    (      BÄ    )      
Å    *      ÒÅ    +      šÆ    ,      bÇ    -      *È    .      òÈ    /      ºÉ    0      ‚Ê    1      JË    2      Ì    3      ÚÌ    4      ¢Í    5      jÎ    6      2Ï    7      úÏ    8      ÂÐ    9      ŠÑ    :      RÒ    ;      Ó    <      âÓ    =      ªÔ    >      rÕ    ?      :Ö    @      ×    A      Êß    B      ’è    C      Zñ    D      "ú    E      êú    F      ²û    G      zü    H      Bý    I      
þ    J      Òþ    K      šÿ    L      b     M      *    N      ò    O      º    P      ‚    Q      J    R          S      Ú    T      ¢    U      j    V      2    W      ú    X      Â	    Y      Š
    Z      R    [          \      â    ]      ª    ^      r    _      :    `          a      Ê    b      ’    c      Z    d      "    e      ê    f      ²    g      z    h      B    i      
    j      Ò    k      š    l      b    m      *    n      ò    o      º    p      ‚    q      J    r          s      Ú    t      ¢    u      j     v      2!    w      ú!    x      Â"    y      Š#    z      R$    {      %    |      â%    }      ª&    ~      r'          :(    €      )    �      Ê1    ‚      ’:    ƒ      ZC    „      "L    …      êL    †      ²M    ‡      zN    ˆ      BO    ‰      
P    Š      ÒP    ‹      šQ    Œ      bR    �      *S    Ž      òS    �      ºT    �      ‚U    ‘      JV    ’      W    “      ÚW    ”      ¢X    •      jY    –      2Z    —      úZ    ˜      Â[    ™      Š\    š      R]    ›      ^    œ      â^    �      ª_    ž      r`    Ÿ      :a           b    ¡      Êb    ¢      ’c    £      Zd    ¤      "e    ¥      êe    ¦      ²f    §      zg    ¨      Bh    ©      
i    ª      Òi    «      šj    ¬      bk    ­      *l    ®      òl    ¯      ºm    °      ‚n    ±      Jo    ²      p    ³      Úp    ´      ¢q    µ      jr    ¶      2s    ·      ús    ¸      Ât    ¹      Šu    º      Rv    »      w    ¼      âw    ½      ªx    ¾      ry    ¿      :z    À      {    Á      Êƒ    Â      ’Œ    Ã      Z•    Ä      "ž    Å      êž    Æ      ²Ÿ    Ç      z     È      B¡    É      
¢    Ê      Ò¢    Ë      š£    Ì      b¤    Í      *¥    Î      ò¥    Ï      º¦    Ð      ‚§    Ñ      J¨    Ò      ©    Ó      Ú©    Ô      ¢ª    Õ      j«    Ö      2¬    ×      ú¬    Ø      Â­    Ù      Š®    Ú      R¯    Û      °    Ü      â°    Ý      ª±    Þ      r²    ß      :³    à      ´    á      Ê´    â      ’µ    ã      Z¶    ä      "·    å      ê·    æ      ²¸    ç      z¹    è      Bº    é      
»    ê      Ò»    ë      š¼    ì      b½    í      *¾    î      ò¾    ï      º¿    ð      ‚À    ñ      JÁ    ò      Â    ó      ÚÂ    ô      ¢Ã    õ      jÄ    ö      2Å    ÷      úÅ    ø      ÂÆ    ù      ŠÇ    ú      RÈ    û      É    ü      âÉ    ý      ªÊ    þ      rË    ÿ      :Ì           Í          ÊÕ          ’Þ          Zç          "ð          êð          ²ñ          zò          Bó    	      
ô    
      Òô          šõ          bö          *÷          ò÷          ºø          ‚ù          Jú          û          Úû          ¢ü          jý          2þ          úþ          Âÿ          Š           R                    â          ª          r          :               !      Ê    "      ’    #      Z    $      "	    %      ê	    &      ²
    '      z    (      B    )      
    *      Ò    +      š    ,      b    -      *    .      ò    /      º    0      ‚    1      J    2          3      Ú    4      ¢    5      j    6      2    7      ú    8      Â    9      Š    :      R    ;          <      â    =      ª    >      r    ?      :    @          A      Ê'    B      ’0    C      Z9    D      "B    E      êB    F      ²C    G      zD    H      BE    I      
F    J      ÒF    K      šG    L      bH    M      *I    N      òI    O      ºJ    P      ‚K    Q      JL    R      M    S      ÚM    T      ¢N    U      jO    V      2P    W      úP    X      ÂQ    Y      ŠR    Z      RS    [      T    \      âT    ]      ªU    ^      rV    _      :W    `      X    a      ÊX    b      ’Y    c      ZZ    d      "[    e      ê[    f      ²\    g      z]    h      B^    i      
_    j      Ò_    k      š`    l      ba    m      *b    n      òb    o      ºc    p      ‚d    q      Je    r      f    s      Úf    t      ¢g    u      jh    v      2i    w      úi    x      Âj    y      Šk    z      Rl    {      m    |      âm    }      ªn    ~      ro          :p    €      q    �      Êy    ‚      ’‚    ƒ      Z‹    „      "”    …      ê”    †      ²•    ‡      z–    ˆ      B—    ‰      
˜    Š      Ò˜    ‹      š™    Œ      bš    �      *›    Ž      ò›    �      ºœ    �      ‚�    ‘      Jž    ’      Ÿ    “      ÚŸ    ”      ¢     •      j¡    –      2¢    —      ú¢    ˜      Â£    ™      Š¤    š      R¥    ›      ¦    œ      â¦    �      ª§    ž      r¨    Ÿ      :©           ª    ¡      Êª    ¢      ’«    £      Z¬    ¤      "­    ¥      ê­    ¦      ²®    §      z¯    ¨      B°    ©      
±    ª      Ò±    «      š²    ¬      b³    ­      *´    ®      ò´    ¯      ºµ    °      ‚¶    ±      J·    ²      ¸    ³      Ú¸    ´      ¢¹    µ      jº    ¶      2»    ·      ú»    ¸      Â¼    ¹      Š½    º      R¾    »      ¿    ¼      â¿    ½      ªÀ    ¾      rÁ    ¿      :Â       ç   8        5   U          9   I          I   Y         U   u          Y   i          q   �            Ÿ          �   ‘          –   ¦         Ÿ   ¿          ¦   ¶         ¿   ß         à                         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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üwu1x”    u1x	4x€    	4xo7xø    o7x[9xŒ    [9x·:xÌ    ·:xç=x€    ç=xÝ>x¨    Ý>xaExÐ    aExõGxä	    õGxJx\    JxiLx0    iLxmNxp    mNxÁ�x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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   §   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üwu1x”    u1x	4x€    	4xo7xø    o7x[9xŒ    [9x±:xÌ    ±:xá=xˆ    á=x×>x°    ×>xaExØ    aExõGxì	    õGxJxd    JxiLx8    iLxsNxx    sNxÇ�x                              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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   ·   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üwu1x”    u1x	4x€    	4xo7xø    o7x[9xŒ    [9x±:xÌ    ±:xŸ@x€    Ÿ@x•Ax¨    •AxaExÐ    aExõGxä	    õGxJx\    JxiLx0    iLxNxp    NxÍ�x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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   §   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üwu1x”    u1x	4x€    	4xu7xø    u7xa9xŒ    a9x±:xÌ    ±:xá=xˆ    á=x×>x°    ×>x=ExØ    =ExÑGxì	    ÑGx½Ixd    ½IxsKx8    sKx…Nxx    …NxÓ�x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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   ·   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ÷   <     áûw×üwp     ×üwu1xœ    u1x	4xˆ    	4xu7x     u7xa9x”    a9x«:xÔ    «:xw;x€    w;x§>x4    §>x�?x\    �?xCEx„	    CEx×Gx˜
    ×Gx·Ix    ·IxmKxä    mKx‹Nx$    ‹Nxß�xœ                             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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   —   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ÷   <     áûw×üwp     ×üwu1xœ    u1x	4xˆ    	4xu7x     u7xa9x”    a9x¥:xÔ    ¥:xË;xˆ    Ë;xû>x<    û>xñ?xd    ñ?xCExŒ	    CEx×Gx 
    ×Gx½Ix    ½IxsKxì    sKx‘Nx,    ‘Nxå�x¤                             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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   §   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ÷   <     áûw×üwp     ×üw{1xœ    {1x4xˆ    4xu7x     u7xa9x”    a9x·:xÔ    ·:xS;xˆ    S;x]@x<    ]@xSAxd    SAxCExŒ	    CEx×Gx 
    ×Gx½Ix    ½IxsKxì    sKx�Nx,    �Nxë�x¤                             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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   §   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üw‡1x”    ‡1x4x€    4x]7xø    ]7xI9xŒ    I9x·:xÌ    ·:xç=xp    ç=xÝ>x˜    Ý>x[ExÀ    [ExïGxÔ	    ïGxÏIxL    ÏIx…Kx     …Kx£Nx`    £Nxñ�x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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   ‡   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üw‡1x”    ‡1x4x€    4x]7xø    ]7xI9xŒ    I9xÉ:xÌ    É:xù=xx    ù=xï>x     ï>x1ExÈ    1ExÅGxÜ	    ÅGx½Ix
    ½IxsKxò     sKx©NxŠ     ©Nxý�x’                              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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   —   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  invocations     1       `        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  ~        ~  †        ‡  �      	   	                    b      ª        %      operation.2     gemm_f16_init_small_rows_interleaved_k16_c16_r18_r18           
     €       &         b      J   ‚          operation.2     exchange                   ½ûwkŒxr        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       []         7            :      z      2            b         reason  OperatorInputs  value   20      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                                              b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üw‡1x”    ‡1x4x€    4x]7xø    ]7xI9xŒ    I9x±:xÌ    ±:x¹Axp    ¹Ax¯Bx˜    ¯BxOExÀ    OExãGxÔ	    ãGxÏIxL    ÏIx…Kx     …Kx¯Nx`    ¯NxŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   ‡   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     áûw×üwh     ×üw�1x”    �1x!4x€    !4x]7xø    ]7xI9xŒ    I9x½:xÌ    ½:xí=xx    í=xã>x     ã>x1ExÈ    1ExÅGxÜ	    ÅGxÃIxT    ÃIxyKx(    yKx»Nxh    »Nx	Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   —   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ½ûwkŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     Ýüw�1x”    �1x!4x€    !4xc7xø    c7xO9xŒ    O9x·:xÌ    ·:xç=xp    ç=xÝ>x˜    Ý>x=ExÀ    =ExÑGxÔ	    ÑGxÕIxL    ÕIx‹Kx     ‹KxÁNx`    ÁNxŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   ‡   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     Ýüw�1x”    �1x!4x€    !4xc7xø    c7xO9xŒ    O9x«:xÌ    «:xÛ=xp    Û=xÑ>x˜    Ñ>x7ExÀ    7ExËGxÔ	    ËGxÏIxL    ÏIx…Kx     …KxÇNx`    ÇNxŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   ‡   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     Ýüw�1x”    �1x!4x€    !4xc7xø    c7xO9xŒ    O9x·:xÌ    ·:x¹Axh    ¹Ax¯Bx�    ¯Bx=Ex¸    =ExÑGxÌ	    ÑGxÏIxD    ÏIx…Kx    …KxÍNxX    ÍNx!Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     Ýüw�1x”    �1x!4x€    !4xc7xø    c7xO9xŒ    O9xÃ:xÌ    Ã:xó=xh    ó=xé>x�    é>x=Ex¸    =ExÑGxÌ	    ÑGxÏIxD    ÏIx…Kx    …KxÙNxX    ÙNx'Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34x]7xø    ]7xI9xŒ    I9x½:xÌ    ½:xí=xh    í=xã>x�    ã>x…Ex¸    …ExHxÌ	    HxáIxD    áIx—Kx    —KxßNxX    ßNx-Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34xc7xø    c7xO9xŒ    O9x½:xÌ    ½:xÅAxh    ÅAx»Bx�    »Bx…Ex¸    …ExHxÌ	    HxÕIxD    ÕIx‹Kx    ‹KxåNxX    åNx9Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34xc7xø    c7xO9xŒ    O9x½:xÌ    ½:xí=xh    í=xã>x�    ã>x…Ex¸    …ExHxÌ	    HxÕIxD    ÕIx‹Kx    ‹KxëNxX    ëNx?Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34xc7xø    c7xO9xŒ    O9xÃ:xÌ    Ã:xó=xh    ó=xé>x�    é>x…Ex¸    …ExHxÌ	    HxÕIxD    ÕIx‹Kx    ‹Kx÷NxX    ÷NxEŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34xc7xø    c7xO9xŒ    O9x·:xÌ    ·:xç=xh    ç=xÝ>x�    Ý>xaEx¸    aExõGxÌ	    õGxÕIxD    ÕIx‹Kx    ‹KxýNxX    ýNxKŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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      Š      7   M      operation.3     repeat-remainder                 :      :      2            Z         reason  Repeat  value   23      iterations      2                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x     Ÿ1x34x     34xc7x
     c7xO9xJ     O9x·:xâ
     ·:x¹Ax2     ¹Ax¯Bxš     ¯BxaEx     aExõGxB     õGxÕIxJ     ÕIx‹Kx
     ‹KxOx¢     OxWŽxª                              b      r               operation.0     fill_zero_u64                     ÃûwqŒxŠ        §   (      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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                                              b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34xi7xø    i7xU9xŒ    U9xÃ:xÌ    Ã:xó=xh    ó=xé>x�    é>xaEx¸    aExõGxÌ	    õGxÛIxD    ÛIx‘Kx    ‘Kx	OxX    	Ox]Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     çûwÝüwh     ÝüwŸ1x”    Ÿ1x34x€    34xi7xø    i7xU9xŒ    U9xÉ:xÌ    É:xù=xh    ù=xï>x�    ï>xaEx¸    aExõGxÌ	    õGxçIxD    çIx�Kx    �KxOxX    OxcŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   w   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÃûwqŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xo7xø    o7x[9xŒ    [9xÃ:xÌ    Ã:xÍ?x`    Í?xÃ@xˆ    Ã@xyEx°    yExHxÄ	    HxÛIx<    ÛIx‘Kx    ‘KxOxP    OxiŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xi7xø    i7xU9xŒ    U9xÉ:xÌ    É:xù=x`    ù=xï>xˆ    ï>xyEx°    yExHxÄ	    HxçIx<    çIx�Kx    �Kx!OxP    !OxuŽx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xo7xø    o7x[9xŒ    [9xÃ:xÌ    Ã:xó=x`    ó=xé>xˆ    é>xyEx°    yExHxÄ	    HxçIx<    çIx�Kx    �Kx'OxP    'Ox{Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xo7xø    o7x[9xŒ    [9xÃ:xÌ    Ã:xó=x`    ó=xé>xˆ    é>xEx°    ExHxÄ	    HxçIx<    çIx�Kx    �Kx3OxP    3Ox�Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                         ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÉûwwŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xo7xø    o7x[9xŒ    [9xÃ:xÌ    Ã:x±@x`    ±@x§Axˆ    §AxmEx°    mExHxÄ	    HxóIx<    óIx©Kx    ©Kx9OxP    9Ox‡Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=xIEx„    IExIxð    IxGMxP    GMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xo7xø    o7x[9xŒ    [9xÉ:xÌ    É:xù=x`    ù=xï>xˆ    ï>xgEx°    gExûGxÄ	    ûGxíIx<    íIx£Kx    £Kx?OxP    ?Ox“Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=x=Ex„    =ExIxð    IxGMxP    GMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ç   8     íûwãüwh     ãüw«1x”    «1x?4x€    ?4xu7xø    u7xa9xŒ    a9xÉ:xÌ    É:xù=x`    ù=xï>xˆ    ï>xmEx°    mExHxÄ	    HxóIx<    óIx©Kx    ©KxEOxP    EOx™Žx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      '   =   7  operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÏ:xè     Ï:xí=x\    í=x1Ex„    1ExIxð    IxGMxP    GMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ÷   <     íûwãüwp     ãüw«1xœ    «1x+3xx    +3x¿5x,    ¿5xu7x¤    u7xa9x8    a9xÏ:xx    Ï:xÿ=x    ÿ=xõ>x4    õ>xmEx\	    mExHxp
    HxóIxè    óIx©Kx¼    ©KxQOxü    QOxŸŽxl                             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      '   =     operation.0     exchange                 :      z      b         reason  OperatorInputs  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   M   g   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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   z   Q   2   Q      Q   Z   U   Â  m   z   q   "   q   j   u      u   š   }      }   ’   …   r  ™   b   �      �   š   ¥   *   ¥   R   ©   :   reason  OperatorInputs  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                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÏ:xè     Ï:xí=x\    í=xEx„    ExIxð    IxGMxP    GMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=xEx„    ExIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÃ:xè     Ã:xá=x\    á=x%Ex„    %ExIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÃ:xè     Ã:xá=x\    á=xãDx„    ãDxIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÃ.x8     Ã.x¯6xˆ     ¯6xÃ:xè     Ã:xá=x\    á=x+Ex„    +ExIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         íûwÉ.x8     É.x¯6xˆ     ¯6xÃ:xè     Ã:xá=x\    á=xEx„    ExIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÉ.x8     É.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=x%Ex„    %ExIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÉ.x8     É.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=xEx„    ExIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         íûwÉ.x8     É.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=xûDx„    ûDxIxð    IxMMxP    MMx›Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÕ:xè     Õ:xó=x\    ó=xEx„    ExIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=x=Ex„    =ExIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÕ:xè     Õ:xó=x\    ó=xéDx„    éDxIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÉ:xè     É:xç=x\    ç=x1Ex„    1ExIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÕ:xè     Õ:xó=x\    ó=x%Ex„    %ExIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÏ:xè     Ï:xí=x\    í=xEx„    ExIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÕ:xè     Õ:xó=x\    ó=xEx„    ExIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.x¯6xˆ     ¯6xÏ:xè     Ï:xí=x\    í=xïDx„    ïDxIxð    IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ïûw}Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÏ:xè     Ï:xí=x\    í=xEx„    Ex	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xá:xè     á:xÿ=x\    ÿ=xEx„    Ex	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÏ:xè     Ï:xí=x\    í=xïDx„    ïDx	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÛ:xè     Û:xù=x\    ù=xEx„    Ex	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÕ:xè     Õ:xó=x\    ó=x+Ex„    +Ex	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÛ:xè     Û:xù=x\    ù=xEx„    Ex	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÕ:xè     Õ:xó=x\    ó=xEx„    Ex	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xÛ:xè     Û:xù=x\    ù=xéDx„    éDx	Ixð    	IxMMxP    MMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                ‡         óûwÉ.x8     É.xµ6xˆ     µ6xá:xè     á:xÿ=x\    ÿ=xEx„    Ex	Ixð    	IxSMxP    SMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xç:xè     ç:x>x\    >x1Ex„    1Ex	Ixð    	IxSMxP    SMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xá:xè     á:xÿ=x\    ÿ=xõDx„    õDx	Ixð    	IxSMxP    SMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                 ‡         óûwÉ.x8     É.xµ6xˆ     µ6xá:xè     á:xÿ=x\    ÿ=x%Ex„    %Ex	Ixð    	IxSMxP    SMx¡Œx°           €                 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   M   '   operation.1     exchange                 :      z      2            b         reason  OperatorInputs  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      '   =   7   operation.2     exchange                 :      z      b         reason  OperatorInputs  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                         ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        ÕûwƒŒx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        óûw¡Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        íûw›Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        çûw•Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        áûw�Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23                        Ûûw‰Œx                                b      *      '   5      operation.3     idle             :      j      2         reason  InactiveTile    value   23             