Update README.md

README.md

@@ -26,6 +26,32 @@ This model is fine-tuned on the SQuAD2.0 dataset. Fine-tuning the biomedical lan
 
 Huggingface library doesn't implement Layer-Wise decay feature, which affects the performance on SQuAD task. The reported result of BioM-ELECTRA-SQuAD in our paper is 88.3 (F1) since we use ELECTRA open-source code with TF checkpoint, which uses Layer-Wise decay. 
 
+
+Training Script
+
+```python
+run_qa.py --model_name_or_path sultan/BioM-ELECTRA-Large-Discriminator \
+--dataset_name squad_v2 \
+--do_train \
+--do_eval \
+--dataloader_num_workers 20 \
+--preprocessing_num_workers 20 \
+--version_2_with_negative \
+--num_train_epochs 2 \
+--learning_rate 5e-5 \
+--max_seq_length 512 \
+--doc_stride 128 \
+--per_device_train_batch_size 8 \
+--gradient_accumulation_steps 6 \
+--per_device_eval_batch_size 128 \
+--fp16 \
+--fp16_opt_level O1 \
+--logging_steps 50 \
+--save_steps 1000 \
+--overwrite_output_dir \
+--output_dir out
+```
+
 Evaluation results on SQuAD2.0 Dev Dataset
 ```
 exact = 84.33420365535248