metadata
license: apache-2.0
library_name: peft
tags:
- trl
- dpo
- generated_from_trainer
base_model: TheBloke/OpenHermes-2-Mistral-7B-GPTQ
model-index:
- name: openhermes-mistral-dpo-gptq
results: []
openhermes-mistral-dpo-gptq
This model is a fine-tuned version of TheBloke/OpenHermes-2-Mistral-7B-GPTQ on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0082
- Rewards/chosen: -0.7999
- Rewards/rejected: -11.8804
- Rewards/accuracies: 1.0
- Rewards/margins: 11.0804
- Logps/rejected: -383.9451
- Logps/chosen: -160.8140
- Logits/rejected: -2.4692
- Logits/chosen: -2.6059
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: 0.0002
- train_batch_size: 1
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 2
- training_steps: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5652 | 0.0 | 10 | 0.4581 | 0.1358 | -0.6615 | 1.0 | 0.7973 | -271.7569 | -151.4569 | -2.4902 | -2.7008 |
| 0.3737 | 0.0 | 20 | 0.1877 | 0.0778 | -2.8831 | 1.0 | 2.9609 | -293.9724 | -152.0366 | -2.4897 | -2.6893 |
| 0.2022 | 0.0 | 30 | 0.0621 | -0.1154 | -6.0503 | 1.0 | 5.9349 | -325.6448 | -153.9687 | -2.4890 | -2.6603 |
| 0.0284 | 0.0 | 40 | 0.0155 | -0.5833 | -10.1231 | 1.0 | 9.5397 | -366.3722 | -158.6483 | -2.4792 | -2.6266 |
| 0.0593 | 0.0 | 50 | 0.0082 | -0.7999 | -11.8804 | 1.0 | 11.0804 | -383.9451 | -160.8140 | -2.4692 | -2.6059 |
Framework versions
- PEFT 0.9.0
- Transformers 4.38.2
- Pytorch 2.0.1+cu117
- Datasets 2.18.0
- Tokenizers 0.15.2