agdsga's picture
update model card README.md
97997ec
metadata
license: apache-2.0
tags:
  - generated_from_trainer
model-index:
  - name: chinese-bert-wwm-finetuned-product-1
    results: []

chinese-bert-wwm-finetuned-product-1

This model is a fine-tuned version of hfl/chinese-bert-wwm on the None dataset. It achieves the following results on the evaluation set:

  • eval_loss: 0.0000
  • eval_runtime: 10.6737
  • eval_samples_per_second: 362.572
  • eval_steps_per_second: 5.715
  • epoch: 11.61
  • step: 18797

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: 2e-05
  • train_batch_size: 256
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20

Framework versions

  • Transformers 4.17.0
  • Pytorch 1.6.0
  • Datasets 2.0.0
  • Tokenizers 0.11.6