Update README.md

README.md

@@ -26,6 +26,26 @@ This model is fine-tuned on the SQuAD2.0 dataset. Fine-tuning the biomedical lan
 
 Huggingface library doesn't implement the Layer-Wise decay feature, which affects the performance on the SQuAD task. The reported result of BioM-ALBERT-xxlarge-SQuAD in our paper is 87.00 (F1) since we use ALBERT open-source code with TF checkpoint, which uses Layer-Wise decay. 
 
+Result with PyTorch and V100 GPU
+
+```
+***** eval metrics *****
+HasAns_exact      = 77.6484
+HasAns_f1         = 85.0136
+HasAns_total      =    5928
+NoAns_exact       =  86.577
+NoAns_f1          =  86.577
+NoAns_total       =    5945
+best_exact        = 82.1191
+best_exact_thresh =     0.0
+best_f1           = 85.7964
+best_f1_thresh    =     0.0
+eval_samples      =   12551
+exact             = 82.1191
+f1                = 85.7964
+total             =   11873
+```
+
 To reproduce results in Google Colab:
 
 - Make sure you have GPU enabled.
@@ -43,13 +63,13 @@ To reproduce results in Google Colab:
 - Run this python code:
 
 ```python
-python /content/transformers/examples/pytorch/question-answering/run_qa.py --model_name_or_path BioM-ALBERT-xxlarge-SQuAD2 \
---do_eval \
---version_2_with_negative \
---per_device_eval_batch_size 8 \
---dataset_name squad_v2 \
---overwrite_output_dir \
---fp16 \
+python /content/transformers/examples/pytorch/question-answering/run_qa.py --model_name_or_path BioM-ALBERT-xxlarge-SQuAD2 \\
+--do_eval \\
+--version_2_with_negative \\
+--per_device_eval_batch_size 8 \\
+--dataset_name squad_v2 \\
+--overwrite_output_dir \\
+--fp16 \\
 --output_dir out
 ```