|
--- |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: REPEAT622_1wangchanberta-base-att-spm-uncased |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# REPEAT622_1wangchanberta-base-att-spm-uncased |
|
|
|
This model is a fine-tuned version of [airesearch/wangchanberta-base-att-spm-uncased](https://huggingface.co/airesearch/wangchanberta-base-att-spm-uncased) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.1400 |
|
|
|
## 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: 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: 5 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|:-------------:|:-----:|:-----:|:---------------:| |
|
| 0.4528 | 1.0 | 9531 | 0.1904 | |
|
| 0.2094 | 2.0 | 19062 | 0.1575 | |
|
| 0.1485 | 3.0 | 28593 | 0.1394 | |
|
| 0.1252 | 4.0 | 38124 | 0.1399 | |
|
| 0.1109 | 5.0 | 47655 | 0.1400 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.15.0 |
|
- Pytorch 1.13.0+cu116 |
|
- Datasets 1.17.0 |
|
- Tokenizers 0.10.3 |
|
|
