Update README.md
Browse files
README.md
CHANGED
@@ -26,6 +26,26 @@ This model is fine-tuned on the SQuAD2.0 dataset. Fine-tuning the biomedical lan
|
|
26 |
|
27 |
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.
|
28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
To reproduce results in Google Colab:
|
30 |
|
31 |
- Make sure you have GPU enabled.
|
@@ -43,13 +63,13 @@ To reproduce results in Google Colab:
|
|
43 |
- Run this python code:
|
44 |
|
45 |
```python
|
46 |
-
python /content/transformers/examples/pytorch/question-answering/run_qa.py --model_name_or_path BioM-ALBERT-xxlarge-SQuAD2
|
47 |
-
--do_eval
|
48 |
-
--version_2_with_negative
|
49 |
-
--per_device_eval_batch_size 8
|
50 |
-
--dataset_name squad_v2
|
51 |
-
--overwrite_output_dir
|
52 |
-
--fp16
|
53 |
--output_dir out
|
54 |
```
|
55 |
|
|
|
26 |
|
27 |
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.
|
28 |
|
29 |
+
Result with PyTorch and V100 GPU
|
30 |
+
|
31 |
+
```
|
32 |
+
***** eval metrics *****
|
33 |
+
HasAns_exact = 77.6484
|
34 |
+
HasAns_f1 = 85.0136
|
35 |
+
HasAns_total = 5928
|
36 |
+
NoAns_exact = 86.577
|
37 |
+
NoAns_f1 = 86.577
|
38 |
+
NoAns_total = 5945
|
39 |
+
best_exact = 82.1191
|
40 |
+
best_exact_thresh = 0.0
|
41 |
+
best_f1 = 85.7964
|
42 |
+
best_f1_thresh = 0.0
|
43 |
+
eval_samples = 12551
|
44 |
+
exact = 82.1191
|
45 |
+
f1 = 85.7964
|
46 |
+
total = 11873
|
47 |
+
```
|
48 |
+
|
49 |
To reproduce results in Google Colab:
|
50 |
|
51 |
- Make sure you have GPU enabled.
|
|
|
63 |
- Run this python code:
|
64 |
|
65 |
```python
|
66 |
+
python /content/transformers/examples/pytorch/question-answering/run_qa.py --model_name_or_path BioM-ALBERT-xxlarge-SQuAD2 \\
|
67 |
+
--do_eval \\
|
68 |
+
--version_2_with_negative \\
|
69 |
+
--per_device_eval_batch_size 8 \\
|
70 |
+
--dataset_name squad_v2 \\
|
71 |
+
--overwrite_output_dir \\
|
72 |
+
--fp16 \\
|
73 |
--output_dir out
|
74 |
```
|
75 |
|