|
--- |
|
base_model: lvwerra/gpt2-imdb |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: gpt-imdb-ipo-beta_0.3 |
|
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. --> |
|
|
|
# gpt-imdb-ipo-beta_0.3 |
|
|
|
This model is a fine-tuned version of [lvwerra/gpt2-imdb](https://huggingface.co/lvwerra/gpt2-imdb) on an unknown dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 1.8601 |
|
- Rewards/chosen: -0.2473 |
|
- Rewards/rejected: -0.6141 |
|
- Rewards/accuracies: 0.8271 |
|
- Rewards/margins: 0.3668 |
|
- Logps/rejected: -265.7321 |
|
- Logps/chosen: -236.0896 |
|
- Logits/rejected: -31.6527 |
|
- Logits/chosen: -31.7977 |
|
|
|
## 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: 1e-05 |
|
- train_batch_size: 24 |
|
- eval_batch_size: 24 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.99) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_steps: 150 |
|
- training_steps: 7197 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| |
|
| 5.822 | 0.21 | 500 | 19.5830 | -0.0268 | -0.3320 | 0.6708 | 0.3052 | -264.7920 | -235.3544 | -33.5002 | -33.8198 | |
|
| 6.8677 | 0.42 | 1000 | 18.7557 | -0.0552 | -0.3293 | 0.5917 | 0.2741 | -264.7829 | -235.4492 | -35.5852 | -35.8178 | |
|
| 12.3698 | 0.63 | 1500 | 36.0453 | -0.1426 | -0.5467 | 0.6771 | 0.4041 | -265.5075 | -235.7406 | -34.3816 | -34.5936 | |
|
| 7.8347 | 0.83 | 2000 | 38.2624 | -0.0799 | -0.3485 | 0.6500 | 0.2687 | -264.8470 | -235.5314 | -33.2874 | -33.4310 | |
|
| 9.184 | 1.04 | 2500 | 14.9546 | -0.3389 | -0.7127 | 0.6875 | 0.3739 | -266.0610 | -236.3948 | -32.7912 | -32.9463 | |
|
| 11.1603 | 1.25 | 3000 | 15.5236 | -0.0513 | -0.3736 | 0.7000 | 0.3223 | -264.9306 | -235.4362 | -33.3399 | -33.5624 | |
|
| 16.5516 | 1.46 | 3500 | 8.6118 | -0.1177 | -0.5526 | 0.7438 | 0.4349 | -265.5274 | -235.6576 | -31.9816 | -32.1630 | |
|
| 5.2761 | 1.67 | 4000 | 5.2168 | -0.1495 | -0.5364 | 0.7417 | 0.3869 | -265.4733 | -235.7637 | -32.2719 | -32.3991 | |
|
| 2.9326 | 1.88 | 4500 | 4.2332 | -0.2284 | -0.6043 | 0.7646 | 0.3759 | -265.6996 | -236.0266 | -32.0240 | -32.1547 | |
|
| 2.9814 | 2.08 | 5000 | 3.3498 | -0.2188 | -0.6063 | 0.7792 | 0.3874 | -265.7062 | -235.9947 | -31.8376 | -31.9728 | |
|
| 1.8651 | 2.29 | 5500 | 2.8900 | -0.2624 | -0.6313 | 0.7896 | 0.3688 | -265.7895 | -236.1400 | -31.4502 | -31.5973 | |
|
| 4.5849 | 2.5 | 6000 | 2.2055 | -0.2771 | -0.6338 | 0.7833 | 0.3567 | -265.7979 | -236.1888 | -31.5011 | -31.6468 | |
|
| 1.7322 | 2.71 | 6500 | 1.9194 | -0.2534 | -0.6145 | 0.8208 | 0.3611 | -265.7336 | -236.1099 | -31.6632 | -31.8054 | |
|
| 1.1697 | 2.92 | 7000 | 1.8601 | -0.2473 | -0.6141 | 0.8271 | 0.3668 | -265.7321 | -236.0896 | -31.6527 | -31.7977 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.35.2 |
|
- Pytorch 2.1.1 |
|
- Datasets 2.15.0 |
|
- Tokenizers 0.15.0 |
|
|
