Edit model card

bert_base_uncased_itr0_0.0001_all_01_03_2022-14_08_15

This model is a fine-tuned version of bert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7632
  • Accuracy: 0.8263
  • F1: 0.8871
  • Precision: 0.8551
  • Recall: 0.9215

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: 0.0001
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Accuracy F1 Precision Recall
No log 1.0 390 0.3986 0.8305 0.8903 0.8868 0.8938
0.4561 2.0 780 0.4018 0.8439 0.9009 0.8805 0.9223
0.3111 3.0 1170 0.4306 0.8354 0.8924 0.8974 0.8875
0.1739 4.0 1560 0.5499 0.8378 0.9002 0.8547 0.9509
0.1739 5.0 1950 0.6223 0.85 0.9052 0.8814 0.9303

Framework versions

  • Transformers 4.15.0
  • Pytorch 1.10.1+cu113
  • Datasets 1.18.0
  • Tokenizers 0.10.3
Downloads last month
4
Inference API
This model can be loaded on Inference API (serverless).