|
|
--- |
|
|
tags: |
|
|
- generated_from_trainer |
|
|
datasets: |
|
|
- korquad |
|
|
model-index: |
|
|
- name: komrc_train |
|
|
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. --> |
|
|
|
|
|
# komrc_train |
|
|
|
|
|
This model is a fine-tuned version of [beomi/kcbert-base](https://huggingface.co/beomi/kcbert-base) on the korquad dataset. |
|
|
It achieves the following results on the evaluation set: |
|
|
- Loss: 0.6544 |
|
|
|
|
|
## 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: 16 |
|
|
- eval_batch_size: 16 |
|
|
- seed: 1234 |
|
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
|
- lr_scheduler_type: linear |
|
|
- num_epochs: 3 |
|
|
- mixed_precision_training: Native AMP |
|
|
|
|
|
### Training results |
|
|
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|
|:-------------:|:-----:|:-----:|:---------------:| |
|
|
| 0.8187 | 0.31 | 2000 | 0.7377 | |
|
|
| 0.6947 | 0.63 | 4000 | 0.6934 | |
|
|
| 0.6352 | 0.94 | 6000 | 0.6544 | |
|
|
| 0.3869 | 1.25 | 8000 | 0.7633 | |
|
|
| 0.3812 | 1.56 | 10000 | 0.7047 | |
|
|
| 0.3579 | 1.88 | 12000 | 0.7097 | |
|
|
| 0.2053 | 2.19 | 14000 | 0.8511 | |
|
|
| 0.2173 | 2.5 | 16000 | 0.8457 | |
|
|
| 0.2094 | 2.82 | 18000 | 0.8433 | |
|
|
|
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- Transformers 4.13.0 |
|
|
- Pytorch 1.10.0+cu111 |
|
|
- Datasets 1.18.4 |
|
|
- Tokenizers 0.10.3 |
|
|
|
