metadata

license: mit
tags:
  - generated_from_trainer
datasets:
  - keyword_pubmed_dataset
metrics:
  - accuracy
model-index:
  - name: kw_pubmed_1000_0.0003
    results:
      - task:
          name: Masked Language Modeling
          type: fill-mask
        dataset:
          name: keyword_pubmed_dataset
          type: keyword_pubmed_dataset
          args: sentence
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.33938523162661094

kw_pubmed_1000_0.0003

This model is a fine-tuned version of microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext on the keyword_pubmed_dataset dataset. It achieves the following results on the evaluation set:

Loss: 4.7086
Accuracy: 0.3394

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: 0.0003
train_batch_size: 32
eval_batch_size: 32
seed: 42
gradient_accumulation_steps: 250
total_train_batch_size: 8000
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 5
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.09	4	4.3723	0.3436
6.0386	0.17	8	4.2113	0.3442
3.7573	0.26	12	4.2079	0.3634
2.9944	0.35	16	4.3370	0.3513
2.7048	0.44	20	4.8594	0.3067
2.7048	0.52	24	4.4929	0.3383
2.9458	0.61	28	4.5146	0.3408
2.3783	0.7	32	4.5680	0.3430
2.2485	0.78	36	4.5095	0.3477
2.1701	0.87	40	4.4971	0.3449
2.1701	0.96	44	4.7051	0.3321
2.0861	1.07	48	4.7615	0.3310
2.4168	1.15	52	4.7086	0.3394

Framework versions

Transformers 4.18.0
Pytorch 1.11.0
Datasets 2.1.0
Tokenizers 0.12.1