hBERTv1_mrpc / README.md
gokuls's picture
End of training
62080f4
metadata
language:
  - en
tags:
  - generated_from_trainer
datasets:
  - glue
metrics:
  - accuracy
  - f1
model-index:
  - name: hBERTv1_mrpc
    results:
      - task:
          name: Text Classification
          type: text-classification
        dataset:
          name: GLUE MRPC
          type: glue
          config: mrpc
          split: validation
          args: mrpc
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.6862745098039216
          - name: F1
            type: f1
            value: 0.7999999999999999

hBERTv1_mrpc

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1 on the GLUE MRPC dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6051
  • Accuracy: 0.6863
  • F1: 0.8000
  • Combined Score: 0.7431

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: 5e-05
  • train_batch_size: 256
  • eval_batch_size: 256
  • seed: 10
  • distributed_type: multi-GPU
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 50
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Accuracy F1 Combined Score
0.6536 1.0 15 0.6243 0.6838 0.8122 0.7480
0.6275 2.0 30 0.6174 0.7010 0.8117 0.7564
0.6129 3.0 45 0.6089 0.6961 0.8182 0.7571
0.6087 4.0 60 0.6062 0.6887 0.8130 0.7508
0.5939 5.0 75 0.6104 0.6863 0.7935 0.7399
0.5707 6.0 90 0.6184 0.7083 0.8183 0.7633
0.5426 7.0 105 0.6051 0.6863 0.8000 0.7431
0.4819 8.0 120 0.6560 0.6936 0.8019 0.7478
0.4279 9.0 135 0.6673 0.6887 0.7678 0.7283
0.3374 10.0 150 0.8092 0.6863 0.7902 0.7382
0.2789 11.0 165 0.9342 0.6887 0.7935 0.7411
0.2216 12.0 180 0.9708 0.6838 0.7810 0.7324

Framework versions

  • Transformers 4.26.1
  • Pytorch 1.14.0a0+410ce96
  • Datasets 2.10.1
  • Tokenizers 0.13.2