metadata
license: apache-2.0
base_model: TheBloke/OpenHermes-2-Mistral-7B-GPTQ
tags:
- generated_from_trainer
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: 2.2177
- Rewards/chosen: -4.0177
- Rewards/rejected: -3.1669
- Rewards/accuracies: 0.5625
- Rewards/margins: -0.8508
- Logps/rejected: -609.4065
- Logps/chosen: -540.1581
- Logits/rejected: -0.8963
- Logits/chosen: -1.0179
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.7105 | 0.01 | 10 | 0.8032 | -0.5208 | -0.1762 | 0.5 | -0.3446 | -579.4993 | -505.1888 | -0.9201 | -1.0572 |
| 0.6608 | 0.01 | 20 | 1.0717 | -1.5841 | -1.1358 | 0.625 | -0.4482 | -589.0955 | -515.8216 | -0.9135 | -1.0449 |
| 0.7477 | 0.01 | 30 | 1.0619 | -1.4672 | -1.4134 | 0.5625 | -0.0539 | -591.8709 | -514.6534 | -0.9162 | -1.0480 |
| 0.7152 | 0.02 | 40 | 1.5518 | -2.5724 | -2.2804 | 0.5625 | -0.2919 | -600.5416 | -525.7047 | -0.9079 | -1.0351 |
| 0.8106 | 0.03 | 50 | 2.2177 | -4.0177 | -3.1669 | 0.5625 | -0.8508 | -609.4065 | -540.1581 | -0.8963 | -1.0179 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.0.1+cu117
- Datasets 2.15.0
- Tokenizers 0.15.0