metadata
tags:
- generated_from_trainer
model-index:
- name: nezha-chinese-base-finetuned-product
results: []
nezha-chinese-base-finetuned-product
This model is a fine-tuned version of peterchou/nezha-chinese-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0004
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: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.0309 | 1.0 | 6473 | 0.0037 |
| 0.0033 | 2.0 | 12946 | 0.0006 |
| 0.0017 | 3.0 | 19419 | 0.0004 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.6.0
- Datasets 2.0.0
- Tokenizers 0.11.6