|
--- |
|
license: llama2 |
|
library_name: peft |
|
tags: |
|
- trl |
|
- dpo |
|
- generated_from_trainer |
|
base_model: meta-llama/Llama-2-7b-hf |
|
model-index: |
|
- name: llama_DPO_model_e2 |
|
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. --> |
|
|
|
# llama_DPO_model_e2 |
|
|
|
This model is a fine-tuned version of [meta-llama/Llama-2-7b-hf](https://huggingface.co/meta-llama/Llama-2-7b-hf) on an unknown dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.0896 |
|
- Rewards/chosen: 0.4401 |
|
- Rewards/rejected: -2.0930 |
|
- Rewards/accuracies: 1.0 |
|
- Rewards/margins: 2.5330 |
|
- Logps/rejected: -205.7391 |
|
- Logps/chosen: -156.2334 |
|
- Logits/rejected: -1.0514 |
|
- Logits/chosen: -0.8587 |
|
|
|
## 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: 8e-07 |
|
- train_batch_size: 1 |
|
- eval_batch_size: 1 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 8 |
|
- total_train_batch_size: 8 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 2 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| |
|
| 0.6699 | 0.1 | 25 | 0.6428 | 0.0307 | -0.0744 | 0.9033 | 0.1051 | -185.5532 | -160.3267 | -1.0520 | -0.8550 | |
|
| 0.5702 | 0.2 | 50 | 0.5471 | 0.0866 | -0.2359 | 0.9933 | 0.3225 | -187.1690 | -159.7680 | -1.0514 | -0.8544 | |
|
| 0.488 | 0.3 | 75 | 0.4456 | 0.1502 | -0.4424 | 1.0 | 0.5926 | -189.2334 | -159.1314 | -1.0527 | -0.8555 | |
|
| 0.3957 | 0.4 | 100 | 0.3600 | 0.2054 | -0.6615 | 1.0 | 0.8669 | -191.4245 | -158.5795 | -1.0530 | -0.8577 | |
|
| 0.3338 | 0.5 | 125 | 0.2865 | 0.2569 | -0.8933 | 1.0 | 1.1502 | -193.7425 | -158.0646 | -1.0524 | -0.8564 | |
|
| 0.253 | 0.6 | 150 | 0.2257 | 0.3043 | -1.1373 | 1.0 | 1.4416 | -196.1830 | -157.5914 | -1.0523 | -0.8570 | |
|
| 0.2134 | 0.7 | 175 | 0.1819 | 0.3496 | -1.3537 | 1.0 | 1.7033 | -198.3466 | -157.1379 | -1.0530 | -0.8584 | |
|
| 0.1613 | 0.79 | 200 | 0.1473 | 0.3842 | -1.5693 | 1.0 | 1.9535 | -200.5027 | -156.7917 | -1.0525 | -0.8591 | |
|
| 0.1358 | 0.89 | 225 | 0.1231 | 0.4031 | -1.7582 | 1.0 | 2.1614 | -202.3919 | -156.6024 | -1.0523 | -0.8593 | |
|
| 0.115 | 0.99 | 250 | 0.1076 | 0.4205 | -1.8980 | 1.0 | 2.3185 | -203.7897 | -156.4292 | -1.0521 | -0.8590 | |
|
| 0.1111 | 1.09 | 275 | 0.0989 | 0.4291 | -1.9856 | 1.0 | 2.4148 | -204.6660 | -156.3426 | -1.0515 | -0.8591 | |
|
| 0.0902 | 1.19 | 300 | 0.0949 | 0.4280 | -2.0337 | 1.0 | 2.4617 | -205.1465 | -156.3540 | -1.0507 | -0.8576 | |
|
| 0.0867 | 1.29 | 325 | 0.0920 | 0.4325 | -2.0705 | 1.0 | 2.5030 | -205.5146 | -156.3087 | -1.0510 | -0.8576 | |
|
| 0.0973 | 1.39 | 350 | 0.0905 | 0.4357 | -2.0839 | 1.0 | 2.5196 | -205.6485 | -156.2766 | -1.0506 | -0.8576 | |
|
| 0.0942 | 1.49 | 375 | 0.0897 | 0.4422 | -2.0838 | 1.0 | 2.5260 | -205.6476 | -156.2122 | -1.0515 | -0.8578 | |
|
| 0.0858 | 1.59 | 400 | 0.0897 | 0.4392 | -2.0903 | 1.0 | 2.5295 | -205.7121 | -156.2415 | -1.0515 | -0.8587 | |
|
| 0.083 | 1.69 | 425 | 0.0893 | 0.4401 | -2.0972 | 1.0 | 2.5373 | -205.7811 | -156.2327 | -1.0511 | -0.8584 | |
|
| 0.0964 | 1.79 | 450 | 0.0897 | 0.4368 | -2.0947 | 1.0 | 2.5315 | -205.7564 | -156.2662 | -1.0511 | -0.8577 | |
|
| 0.0931 | 1.89 | 475 | 0.0890 | 0.4406 | -2.0970 | 1.0 | 2.5376 | -205.7794 | -156.2282 | -1.0512 | -0.8585 | |
|
| 0.0915 | 1.99 | 500 | 0.0896 | 0.4401 | -2.0930 | 1.0 | 2.5330 | -205.7391 | -156.2334 | -1.0514 | -0.8587 | |
|
|
|
|
|
### Framework versions |
|
|
|
- PEFT 0.8.2 |
|
- Transformers 4.38.1 |
|
- Pytorch 2.2.0+cu118 |
|
- Datasets 2.17.1 |
|
- Tokenizers 0.15.2 |