Bert_Classifier / README.md
jakka's picture
update model card README.md
4501fb0
metadata
license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - yelp_review_full
metrics:
  - accuracy
model-index:
  - name: Bert_Classifier
    results:
      - task:
          name: Text Classification
          type: text-classification
        dataset:
          name: yelp_review_full
          type: yelp_review_full
          config: yelp_review_full
          split: train
          args: yelp_review_full
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.43

Bert_Classifier

This model is a fine-tuned version of bert-base-cased on the yelp_review_full dataset. It achieves the following results on the evaluation set:

  • Loss: 1.9851
  • Accuracy: 0.43

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: 8
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Accuracy
No log 1.0 125 1.5585 0.45
No log 2.0 250 1.7005 0.51
No log 3.0 375 1.9851 0.43

Framework versions

  • Transformers 4.21.3
  • Pytorch 1.12.1+cu113
  • Datasets 2.4.0
  • Tokenizers 0.12.1