CRAFT_PubMedBERT_NER

• Loss: 0.1043
• Seqeval classification report: precision recall f1-score support

   CHEBI       0.71      0.73      0.72       616
      CL       0.85      0.89      0.87      1740
     GGP       0.84      0.76      0.80       611
      GO       0.89      0.90      0.90      3810
      SO       0.81      0.83      0.82      8854
   Taxon       0.58      0.60      0.59       284

  micro avg 0.82 0.84 0.83 15915 macro avg 0.78 0.79 0.78 15915 weighted avg 0.82 0.84 0.83 15915

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• 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       0.66      0.61      0.63       616
      CL       0.81      0.86      0.83      1740
     GGP       0.74      0.54      0.63       611
      GO       0.86      0.89      0.87      3810
      SO       0.73      0.78      0.76      8854
   Taxon       0.47      0.57      0.52       284

   CHEBI       0.69      0.74      0.71       616
      CL       0.84      0.88      0.86      1740
     GGP       0.83      0.74      0.78       611
      GO       0.88      0.90      0.89      3810
      SO       0.79      0.82      0.81      8854
   Taxon       0.57      0.60      0.58       284

   CHEBI       0.71      0.73      0.72       616
      CL       0.85      0.89      0.87      1740
     GGP       0.84      0.76      0.80       611
      GO       0.89      0.90      0.90      3810
      SO       0.81      0.83      0.82      8854
   Taxon       0.58      0.60      0.59       284

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0