Update README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
-
BioM-Transformers: Building Large Biomedical Language Models with BERT, ALBERT and ELECTRA
|
2 |
|
3 |
-
Abstract
|
4 |
|
5 |
|
6 |
The impact of design choices on the performance
|
@@ -20,6 +20,8 @@ the significant effect of design choices on
|
|
20 |
improving the performance of biomedical language
|
21 |
models.
|
22 |
|
|
|
|
|
23 |
This model is fine-tuned on the SQuAD2.0 dataset. Fine-tuning the biomedical language model on the SQuAD dataset helps improve the score on the BioASQ challenge. If you plan to work with BioASQ or biomedical QA tasks, it's better to use this model over BioM-ELECTRA-Large. This model (TensorFlow version ) took the lead in the BioASQ9b-Factoid challenge under the name of (UDEL-LAB2). To see the full details of BioASQ9B results, please check this link http://participants-area.bioasq.org/results/9b/phaseB/ ( you need to register).
|
24 |
|
25 |
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.
|
@@ -58,13 +60,13 @@ To reproduce results in Google Colab:
|
|
58 |
- Run this python code:
|
59 |
|
60 |
```python
|
61 |
-
python /content/transformers/examples/pytorch/question-answering/run_qa.py --model_name_or_path sultan/BioM-ELECTRA-Large-SQuAD2
|
62 |
-
--do_eval
|
63 |
-
--version_2_with_negative
|
64 |
-
--per_device_eval_batch_size 8
|
65 |
-
--dataset_name squad_v2
|
66 |
-
--overwrite_output_dir
|
67 |
-
--fp16
|
68 |
--output_dir out
|
69 |
```
|
70 |
|
@@ -73,11 +75,11 @@ You don't need to download the SQuAD2 dataset. The code will download it from th
|
|
73 |
|
74 |
Check our GitHub repo at https://github.com/salrowili/BioM-Transformers for TensorFlow and GluonNLP checkpoints.
|
75 |
|
76 |
-
Acknowledgment
|
77 |
|
78 |
We would like to acknowledge the support we have from Tensorflow Research Cloud (TFRC) team to grant us access to TPUv3 units.
|
79 |
|
80 |
-
Citation
|
81 |
|
82 |
|
83 |
```bibtex
|
|
|
1 |
+
# BioM-Transformers: Building Large Biomedical Language Models with BERT, ALBERT and ELECTRA
|
2 |
|
3 |
+
#Abstract
|
4 |
|
5 |
|
6 |
The impact of design choices on the performance
|
|
|
20 |
improving the performance of biomedical language
|
21 |
models.
|
22 |
|
23 |
+
# Model Description
|
24 |
+
|
25 |
This model is fine-tuned on the SQuAD2.0 dataset. Fine-tuning the biomedical language model on the SQuAD dataset helps improve the score on the BioASQ challenge. If you plan to work with BioASQ or biomedical QA tasks, it's better to use this model over BioM-ELECTRA-Large. This model (TensorFlow version ) took the lead in the BioASQ9b-Factoid challenge under the name of (UDEL-LAB2). To see the full details of BioASQ9B results, please check this link http://participants-area.bioasq.org/results/9b/phaseB/ ( you need to register).
|
26 |
|
27 |
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.
|
|
|
60 |
- Run this python code:
|
61 |
|
62 |
```python
|
63 |
+
python /content/transformers/examples/pytorch/question-answering/run_qa.py --model_name_or_path sultan/BioM-ELECTRA-Large-SQuAD2 \\\\\\\\\\\\\\\\
|
64 |
+
--do_eval \\\\\\\\\\\\\\\\
|
65 |
+
--version_2_with_negative \\\\\\\\\\\\\\\\
|
66 |
+
--per_device_eval_batch_size 8 \\\\\\\\\\\\\\\\
|
67 |
+
--dataset_name squad_v2 \\\\\\\\\\\\\\\\
|
68 |
+
--overwrite_output_dir \\\\\\\\\\\\\\\\
|
69 |
+
--fp16 \\\\\\\\\\\\\\\\
|
70 |
--output_dir out
|
71 |
```
|
72 |
|
|
|
75 |
|
76 |
Check our GitHub repo at https://github.com/salrowili/BioM-Transformers for TensorFlow and GluonNLP checkpoints.
|
77 |
|
78 |
+
# Acknowledgment
|
79 |
|
80 |
We would like to acknowledge the support we have from Tensorflow Research Cloud (TFRC) team to grant us access to TPUv3 units.
|
81 |
|
82 |
+
# Citation
|
83 |
|
84 |
|
85 |
```bibtex
|