ckpts/llama2-13b-viettel_v3.2_1epoch
This model is a fine-tuned version of meta-llama/Llama-2-13b-hf on the our custom dataset. It achieves the following results on the evaluation set:
- Loss: 0.3534
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- total_eval_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 20
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.4028 | 0.08 | 200 | 0.3990 |
| 0.3973 | 0.16 | 400 | 0.3866 |
| 0.3832 | 0.24 | 600 | 0.3790 |
| 0.3844 | 0.33 | 800 | 0.3728 |
| 0.3703 | 0.41 | 1000 | 0.3676 |
| 0.3682 | 0.49 | 1200 | 0.3640 |
| 0.3669 | 0.57 | 1400 | 0.3606 |
| 0.3677 | 0.65 | 1600 | 0.3580 |
| 0.3545 | 0.73 | 1800 | 0.3556 |
| 0.3593 | 0.82 | 2000 | 0.3543 |
| 0.3442 | 0.9 | 2200 | 0.3536 |
| 0.363 | 0.98 | 2400 | 0.3534 |
Framework versions
- Transformers 4.34.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.14.0