Update README.md
Browse files
README.md
CHANGED
|
@@ -24,12 +24,12 @@ model_index:
|
|
| 24 |
value: 0.8640646029609691
|
| 25 |
---
|
| 26 |
|
| 27 |
-
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 28 |
-
should probably proofread and complete it, then remove this comment. -->
|
| 29 |
-
|
| 30 |
# BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext-finetuned-renet
|
| 31 |
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
| 33 |
It achieves the following results on the evaluation set:
|
| 34 |
- Loss: 0.7226
|
| 35 |
- Precision: 0.7799
|
|
@@ -38,19 +38,8 @@ It achieves the following results on the evaluation set:
|
|
| 38 |
- Accuracy: 0.8641
|
| 39 |
- Auc: 0.9325
|
| 40 |
|
| 41 |
-
## Model description
|
| 42 |
-
|
| 43 |
-
More information needed
|
| 44 |
-
|
| 45 |
-
## Intended uses & limitations
|
| 46 |
-
|
| 47 |
-
More information needed
|
| 48 |
-
|
| 49 |
-
## Training and evaluation data
|
| 50 |
-
|
| 51 |
-
More information needed
|
| 52 |
-
|
| 53 |
## Training procedure
|
|
|
|
| 54 |
|
| 55 |
### Training hyperparameters
|
| 56 |
|
|
@@ -63,9 +52,6 @@ The following hyperparameters were used during training:
|
|
| 63 |
- lr_scheduler_type: linear
|
| 64 |
- num_epochs: 5
|
| 65 |
|
| 66 |
-
### Training results
|
| 67 |
-
|
| 68 |
-
|
| 69 |
|
| 70 |
### Framework versions
|
| 71 |
|
|
|
|
| 24 |
value: 0.8640646029609691
|
| 25 |
---
|
| 26 |
|
|
|
|
|
|
|
|
|
|
| 27 |
# BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext-finetuned-renet
|
| 28 |
|
| 29 |
+
A model for detecting gene disease associations from abstracts. The model classifies as 0 for no association, or 1 for some association.
|
| 30 |
+
|
| 31 |
+
This model is a fine-tuned version of [microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext](https://huggingface.co/microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext) on the [RENET2](https://github.com/sujunhao/RENET2) dataset. Note that this considers only the abstract data, and not the full text information, from RENET2.
|
| 32 |
+
|
| 33 |
It achieves the following results on the evaluation set:
|
| 34 |
- Loss: 0.7226
|
| 35 |
- Precision: 0.7799
|
|
|
|
| 38 |
- Accuracy: 0.8641
|
| 39 |
- Auc: 0.9325
|
| 40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
## Training procedure
|
| 42 |
+
The abstract dataset from RENET2 was split into 85% train, 15% evaluation being grouped by PMIDs and stratified by labels. That is, no data from the same PMID was seen in multiple both the training and the evaluation set.
|
| 43 |
|
| 44 |
### Training hyperparameters
|
| 45 |
|
|
|
|
| 52 |
- lr_scheduler_type: linear
|
| 53 |
- num_epochs: 5
|
| 54 |
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
### Framework versions
|
| 57 |
|