0ae96ff8a7
* Created using Colaboratory * [examples] reorganize files * remove run_tpu_glue.py as superseded by TPU support in Trainer * Bugfix: int, not tuple * move files around
32 lines
792 B
Markdown
32 lines
792 B
Markdown
## Multiple Choice
|
|
|
|
Based on the script [`run_multiple_choice.py`]().
|
|
|
|
#### Fine-tuning on SWAG
|
|
Download [swag](https://github.com/rowanz/swagaf/tree/master/data) data
|
|
|
|
```bash
|
|
#training on 4 tesla V100(16GB) GPUS
|
|
export SWAG_DIR=/path/to/swag_data_dir
|
|
python ./examples/run_multiple_choice.py \
|
|
--task_name swag \
|
|
--model_name_or_path roberta-base \
|
|
--do_train \
|
|
--do_eval \
|
|
--data_dir $SWAG_DIR \
|
|
--learning_rate 5e-5 \
|
|
--num_train_epochs 3 \
|
|
--max_seq_length 80 \
|
|
--output_dir models_bert/swag_base \
|
|
--per_gpu_eval_batch_size=16 \
|
|
--per_gpu_train_batch_size=16 \
|
|
--gradient_accumulation_steps 2 \
|
|
--overwrite_output
|
|
```
|
|
Training with the defined hyper-parameters yields the following results:
|
|
```
|
|
***** Eval results *****
|
|
eval_acc = 0.8338998300509847
|
|
eval_loss = 0.44457291918821606
|
|
```
|