|
--- |
|
base_model: lvwerra/gpt2-imdb |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: gpt-imdb-ipo-beta_0.1 |
|
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_0.1 |
|
|
|
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: |
|
- Step: 6500 |
|
- Loss: 11.7007 |
|
- Rewards/chosen: -0.0805 |
|
- Rewards/rejected: -0.4417 |
|
- Rewards/accuracies: 0.9000 |
|
- Rewards/margins: 0.3612 |
|
- Logps/rejected: -268.1027 |
|
- Logps/chosen: -236.0704 |
|
- Logits/rejected: -31.0790 |
|
- Logits/chosen: -31.2840 |
|
|
|
## 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 | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| |
|
| 18.812 | 0.21 | 500 | 29.2155 | 0.0458 | -0.2317 | 0.7875 | 0.2775 | -266.0027 | -234.8074 | -33.9160 | -34.3504 | |
|
| 13.7881 | 0.42 | 1000 | 24.1460 | -0.0697 | -0.3582 | 0.7625 | 0.2885 | -267.2670 | -235.9622 | -35.0526 | -35.3757 | |
|
| 27.0047 | 0.63 | 1500 | 39.7182 | -0.1370 | -0.4692 | 0.7875 | 0.3322 | -268.3775 | -236.6354 | -32.1933 | -32.4137 | |
|
| 19.7751 | 0.83 | 2000 | 40.6223 | -0.0674 | -0.4210 | 0.7729 | 0.3536 | -267.8954 | -235.9392 | -31.7349 | -31.9095 | |
|
| 9.5381 | 1.04 | 2500 | 20.9269 | -0.1155 | -0.4866 | 0.8146 | 0.3712 | -268.5513 | -236.4198 | -32.1382 | -32.3448 | |
|
| 20.3498 | 1.25 | 3000 | 29.2158 | -0.0629 | -0.4040 | 0.8208 | 0.3410 | -267.7249 | -235.8945 | -31.7900 | -32.1080 | |
|
| 20.4018 | 1.46 | 3500 | 20.8452 | -0.0350 | -0.3582 | 0.8271 | 0.3232 | -267.2670 | -235.6155 | -31.3911 | -31.6578 | |
|
| 17.4506 | 1.67 | 4000 | 16.4207 | -0.1258 | -0.4841 | 0.8438 | 0.3583 | -268.5259 | -236.5234 | -31.5718 | -31.7727 | |
|
| 7.7045 | 1.88 | 4500 | 14.3286 | -0.0659 | -0.4275 | 0.875 | 0.3616 | -267.9600 | -235.9239 | -31.3055 | -31.4702 | |
|
| 9.4274 | 2.08 | 5000 | 12.6249 | -0.1037 | -0.4565 | 0.8687 | 0.3528 | -268.2499 | -236.3019 | -31.4025 | -31.6122 | |
|
| 7.7699 | 2.29 | 5500 | 12.3366 | -0.0787 | -0.4337 | 0.8708 | 0.3550 | -268.0224 | -236.0526 | -30.8436 | -31.0563 | |
|
| 9.2038 | 2.5 | 6000 | 12.2158 | -0.0882 | -0.4430 | 0.8937 | 0.3548 | -268.1148 | -236.1471 | -30.7819 | -30.9884 | |
|
| 11.4596 | 2.71 | 6500 | 11.7007 | -0.0852 | -0.4480 | 0.9000 | 0.3628 | -268.1655 | -236.1172 | -31.0236 | -31.2283 | |
|
| 9.6351 | 2.92 | 7000 | 12.0082 | -0.0805 | -0.4417 | 0.8958 | 0.3612 | -268.1027 | -236.0704 | -31.0790 | -31.2840 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.35.2 |
|
- Pytorch 2.1.1 |
|
- Datasets 2.15.0 |
|
- Tokenizers 0.15.0 |
|
|