judithrosell / CRAFT_ClinicalBERT_NER

CRAFT_ClinicalBERT_NER

This model is a fine-tuned version of medicalai/ClinicalBERT on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1733

• Seqeval classification report: precision recall f1-score support

   CHEBI       0.68      0.66      0.67      1365
      CL       0.55      0.50      0.52       284
     GGP       0.87      0.81      0.84      4632
      GO       0.66      0.65      0.65      8852
      SO       0.68      0.50      0.58       616
   Taxon       0.81      0.73      0.77       986
  

  micro avg 0.72 0.69 0.71 16735 macro avg 0.71 0.64 0.67 16735

weighted avg 0.73 0.69 0.71 16735

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

   CHEBI       0.64      0.56      0.60      1365
      CL       0.53      0.35      0.42       284
     GGP       0.84      0.77      0.81      4632
      GO       0.60      0.61      0.60      8852
      SO       0.53      0.46      0.49       616
   Taxon       0.78      0.66      0.71       986

micro avg 0.68 0.64 0.66 16735 macro avg 0.65 0.57 0.61 16735 weighted avg 0.68 0.64 0.66 16735 | | 0.2231 | 2.0 | 695 | 0.1740 | precision recall f1-score support

   CHEBI       0.69      0.63      0.66      1365
      CL       0.56      0.44      0.49       284
     GGP       0.83      0.79      0.81      4632
      GO       0.65      0.65      0.65      8852
      SO       0.68      0.47      0.55       616
   Taxon       0.81      0.72      0.76       986

micro avg 0.71 0.68 0.69 16735 macro avg 0.70 0.62 0.65 16735 weighted avg 0.71 0.68 0.69 16735 | | 0.0813 | 3.0 | 1041 | 0.1733 | precision recall f1-score support

   CHEBI       0.68      0.66      0.67      1365
      CL       0.55      0.50      0.52       284
     GGP       0.87      0.81      0.84      4632
      GO       0.66      0.65      0.65      8852
      SO       0.68      0.50      0.58       616
   Taxon       0.81      0.73      0.77       986

micro avg 0.72 0.69 0.71 16735 macro avg 0.71 0.64 0.67 16735 weighted avg 0.73 0.69 0.71 16735 |

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0