|
--- |
|
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: 0.1526 |
|
- Rewards/chosen: 0.3611 |
|
- Rewards/rejected: -1.5450 |
|
- Rewards/accuracies: 1.0 |
|
- Rewards/margins: 1.9061 |
|
- Logps/rejected: -200.2592 |
|
- Logps/chosen: -157.0226 |
|
- Logits/rejected: -1.0513 |
|
- Logits/chosen: -0.8571 |
|
|
|
## 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: 5e-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: 3 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| |
|
| 0.6819 | 0.1 | 25 | 0.6708 | 0.0151 | -0.0312 | 0.7567 | 0.0463 | -185.1220 | -160.4831 | -1.0517 | -0.8540 | |
|
| 0.6351 | 0.2 | 50 | 0.6228 | 0.0428 | -0.1054 | 0.9600 | 0.1482 | -185.8636 | -160.2060 | -1.0524 | -0.8552 | |
|
| 0.5874 | 0.3 | 75 | 0.5655 | 0.0762 | -0.2019 | 0.9967 | 0.2781 | -186.8286 | -159.8719 | -1.0525 | -0.8548 | |
|
| 0.5179 | 0.4 | 100 | 0.5030 | 0.1133 | -0.3207 | 1.0 | 0.4340 | -188.0166 | -159.5010 | -1.0521 | -0.8545 | |
|
| 0.479 | 0.5 | 125 | 0.4468 | 0.1501 | -0.4388 | 1.0 | 0.5889 | -189.1974 | -159.1327 | -1.0524 | -0.8554 | |
|
| 0.406 | 0.6 | 150 | 0.3904 | 0.1842 | -0.5778 | 1.0 | 0.7620 | -190.5874 | -158.7915 | -1.0525 | -0.8576 | |
|
| 0.3731 | 0.7 | 175 | 0.3377 | 0.2223 | -0.7247 | 1.0 | 0.9470 | -192.0564 | -158.4104 | -1.0521 | -0.8559 | |
|
| 0.3075 | 0.79 | 200 | 0.2918 | 0.2537 | -0.8769 | 1.0 | 1.1305 | -193.5782 | -158.0974 | -1.0525 | -0.8583 | |
|
| 0.2621 | 0.89 | 225 | 0.2517 | 0.2822 | -1.0278 | 1.0 | 1.3100 | -195.0876 | -157.8119 | -1.0525 | -0.8573 | |
|
| 0.2285 | 0.99 | 250 | 0.2180 | 0.3118 | -1.1738 | 1.0 | 1.4855 | -196.5471 | -157.5160 | -1.0517 | -0.8568 | |
|
| 0.2162 | 1.09 | 275 | 0.1948 | 0.3279 | -1.2897 | 1.0 | 1.6176 | -197.7066 | -157.3551 | -1.0513 | -0.8567 | |
|
| 0.1752 | 1.19 | 300 | 0.1810 | 0.3383 | -1.3661 | 1.0 | 1.7044 | -198.4706 | -157.2514 | -1.0511 | -0.8576 | |
|
| 0.1672 | 1.29 | 325 | 0.1714 | 0.3456 | -1.4242 | 1.0 | 1.7698 | -199.0516 | -157.1775 | -1.0509 | -0.8568 | |
|
| 0.1722 | 1.39 | 350 | 0.1646 | 0.3535 | -1.4653 | 1.0 | 1.8187 | -199.4624 | -157.0993 | -1.0510 | -0.8568 | |
|
| 0.1649 | 1.49 | 375 | 0.1596 | 0.3586 | -1.4919 | 1.0 | 1.8505 | -199.7286 | -157.0477 | -1.0512 | -0.8569 | |
|
| 0.1534 | 1.59 | 400 | 0.1580 | 0.3603 | -1.5059 | 1.0 | 1.8663 | -199.8687 | -157.0304 | -1.0507 | -0.8571 | |
|
| 0.1492 | 1.69 | 425 | 0.1561 | 0.3589 | -1.5194 | 1.0 | 1.8783 | -200.0034 | -157.0448 | -1.0514 | -0.8578 | |
|
| 0.1625 | 1.79 | 450 | 0.1564 | 0.3586 | -1.5205 | 1.0 | 1.8791 | -200.0150 | -157.0482 | -1.0509 | -0.8570 | |
|
| 0.1561 | 1.89 | 475 | 0.1535 | 0.3613 | -1.5366 | 1.0 | 1.8979 | -200.1756 | -157.0212 | -1.0510 | -0.8576 | |
|
| 0.1565 | 1.99 | 500 | 0.1529 | 0.3643 | -1.5393 | 1.0 | 1.9036 | -200.2028 | -156.9913 | -1.0513 | -0.8567 | |
|
| 0.1476 | 2.09 | 525 | 0.1530 | 0.3640 | -1.5392 | 1.0 | 1.9032 | -200.2021 | -156.9944 | -1.0511 | -0.8569 | |
|
| 0.1457 | 2.19 | 550 | 0.1530 | 0.3605 | -1.5406 | 1.0 | 1.9011 | -200.2155 | -157.0287 | -1.0507 | -0.8577 | |
|
| 0.1376 | 2.29 | 575 | 0.1529 | 0.3585 | -1.5466 | 1.0 | 1.9051 | -200.2757 | -157.0492 | -1.0508 | -0.8579 | |
|
| 0.1574 | 2.38 | 600 | 0.1527 | 0.3634 | -1.5448 | 1.0 | 1.9082 | -200.2574 | -156.9998 | -1.0508 | -0.8566 | |
|
| 0.1662 | 2.48 | 625 | 0.1518 | 0.3645 | -1.5465 | 1.0 | 1.9109 | -200.2742 | -156.9890 | -1.0509 | -0.8572 | |
|
| 0.1535 | 2.58 | 650 | 0.1523 | 0.3628 | -1.5458 | 1.0 | 1.9086 | -200.2675 | -157.0059 | -1.0510 | -0.8571 | |
|
| 0.1488 | 2.68 | 675 | 0.1518 | 0.3658 | -1.5446 | 1.0 | 1.9104 | -200.2561 | -156.9763 | -1.0510 | -0.8572 | |
|
| 0.1564 | 2.78 | 700 | 0.1526 | 0.3618 | -1.5452 | 1.0 | 1.9071 | -200.2618 | -157.0154 | -1.0512 | -0.8568 | |
|
| 0.1367 | 2.88 | 725 | 0.1526 | 0.3643 | -1.5426 | 1.0 | 1.9069 | -200.2352 | -156.9905 | -1.0513 | -0.8570 | |
|
| 0.1543 | 2.98 | 750 | 0.1526 | 0.3611 | -1.5450 | 1.0 | 1.9061 | -200.2592 | -157.0226 | -1.0513 | -0.8571 | |
|
|
|
|
|
### Framework versions |
|
|
|
- PEFT 0.8.2 |
|
- Transformers 4.38.1 |
|
- Pytorch 2.2.0+cu118 |
|
- Datasets 2.17.1 |
|
- Tokenizers 0.15.2 |