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: 0.6066
- Rewards/chosen: 0.0894
- Rewards/rejected: -0.0023
- Rewards/accuracies: 0.5
- Rewards/margins: 0.0917
- Logps/rejected: -158.0340
- Logps/chosen: -152.2324
- Logits/rejected: -2.3521
- Logits/chosen: -2.4734
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6778 | 0.01 | 10 | 0.6796 | -0.0179 | -0.0353 | 0.625 | 0.0174 | -158.3634 | -153.3049 | -2.3602 | -2.4849 |
| 0.7178 | 0.01 | 20 | 0.6321 | 0.0123 | -0.0701 | 0.5625 | 0.0824 | -158.7113 | -153.0028 | -2.3602 | -2.4815 |
| 0.7263 | 0.01 | 30 | 0.6050 | 0.0538 | -0.0401 | 0.5 | 0.0939 | -158.4116 | -152.5882 | -2.3567 | -2.4779 |
| 0.6941 | 0.02 | 40 | 0.6047 | 0.0834 | -0.0114 | 0.5 | 0.0948 | -158.1251 | -152.2924 | -2.3545 | -2.4749 |
| 0.6889 | 0.03 | 50 | 0.6066 | 0.0894 | -0.0023 | 0.5 | 0.0917 | -158.0340 | -152.2324 | -2.3521 | -2.4734 |
Framework versions
- PEFT 0.9.0
- Transformers 4.38.2
- Pytorch 2.0.1+cu117
- Datasets 2.18.0
- Tokenizers 0.15.2