# Multiple Choice ## Fine-tuning on SWAG with the Trainer `run_swag` allows you to fine-tune any model from our [hub](https://huggingface.co/models) (as long as its architecture as a `ForMultipleChoice` version in the library) on the SWAG dataset or your own csv/jsonlines files as long as they are structured the same way. To make it works on another dataset, you will need to tweak the `preprocess_function` inside the script. ``` python examples/multiple-choice/run_swag.py \ --model_name_or_path roberta-base \ --ipu_config_name Graphcore/roberta-base-ipu \ --do_train \ --do_eval \ --learning_rate 5e-5 \ --num_train_epochs 3 \ --output_dir ./output/swag_base \ --per_device_eval_batch_size=2 \ --per_device_train_batch_size=2 \ --gradient_accumulation_steps 16 \ --n_ipu 16 \ --report_to none \ --pad_on_batch_axis \ --overwrite_output ``` Training with the defined hyper-parameters yields the following results: ``` ***** Eval results ***** eval_acc = 0.8396 eval_loss = 0.439 ```