<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
biobert-base-cased-v1.2-finetuned-ner-Concat_CRAFT_es
This model is a fine-tuned version of dmis-lab/biobert-base-cased-v1.2 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.2079
- Precision: 0.8487
- Recall: 0.8443
- F1: 0.8465
- Accuracy: 0.9693
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 4
Training results
Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|
0.0698 | 1.0 | 2719 | 0.1463 | 0.8132 | 0.8233 | 0.8182 | 0.9643 |
0.0321 | 2.0 | 5438 | 0.1612 | 0.8321 | 0.8463 | 0.8392 | 0.9681 |
0.0154 | 3.0 | 8157 | 0.1832 | 0.8448 | 0.8404 | 0.8426 | 0.9683 |
0.0058 | 4.0 | 10876 | 0.2079 | 0.8487 | 0.8443 | 0.8465 | 0.9693 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.4
- Tokenizers 0.11.6