| # BERT-Small CORD-19 fine-tuned on SQuAD 2.0 | |
| [bert-small-cord19 model](https://huggingface.co/NeuML/bert-small-cord19) fine-tuned on SQuAD 2.0 | |
| ## Building the model | |
| ```bash | |
| python run_squad.py | |
| --model_type bert | |
| --model_name_or_path bert-small-cord19 | |
| --do_train | |
| --do_eval | |
| --do_lower_case | |
| --version_2_with_negative | |
| --train_file train-v2.0.json | |
| --predict_file dev-v2.0.json | |
| --per_gpu_train_batch_size 8 | |
| --learning_rate 3e-5 | |
| --num_train_epochs 3.0 | |
| --max_seq_length 384 | |
| --doc_stride 128 | |
| --output_dir bert-small-cord19-squad2 | |
| --save_steps 0 | |
| --threads 8 | |
| --overwrite_cache | |
| --overwrite_output_dir | |