model_hh_usp3_dpo1
This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.9835
- Rewards/chosen: -10.6241
- Rewards/rejected: -14.4122
- Rewards/accuracies: 0.7300
- Rewards/margins: 3.7881
- Logps/rejected: -258.5334
- Logps/chosen: -217.0869
- Logits/rejected: -0.8154
- Logits/chosen: -0.8130
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1146 | 2.67 | 100 | 0.6900 | -4.8037 | -6.4467 | 0.7300 | 1.6430 | -178.8784 | -158.8832 | -0.8887 | -0.9297 |
| 0.0024 | 5.33 | 200 | 1.0780 | -9.2061 | -12.3558 | 0.7000 | 3.1497 | -237.9695 | -202.9073 | -0.8656 | -0.8655 |
| 0.0001 | 8.0 | 300 | 0.9490 | -10.1793 | -13.8565 | 0.7200 | 3.6772 | -252.9766 | -212.6395 | -0.8313 | -0.8261 |
| 0.0001 | 10.67 | 400 | 0.9700 | -10.4127 | -14.1403 | 0.7300 | 3.7276 | -255.8140 | -214.9731 | -0.8237 | -0.8199 |
| 0.0001 | 13.33 | 500 | 0.9721 | -10.5124 | -14.2790 | 0.7300 | 3.7666 | -257.2012 | -215.9702 | -0.8195 | -0.8168 |
| 0.0001 | 16.0 | 600 | 0.9839 | -10.5785 | -14.3606 | 0.7200 | 3.7820 | -258.0171 | -216.6317 | -0.8172 | -0.8146 |
| 0.0001 | 18.67 | 700 | 0.9829 | -10.6126 | -14.4106 | 0.7300 | 3.7980 | -258.5171 | -216.9720 | -0.8160 | -0.8135 |
| 0.0001 | 21.33 | 800 | 0.9837 | -10.6177 | -14.4091 | 0.7200 | 3.7913 | -258.5019 | -217.0236 | -0.8153 | -0.8132 |
| 0.0001 | 24.0 | 900 | 0.9832 | -10.6228 | -14.4149 | 0.7300 | 3.7921 | -258.5607 | -217.0746 | -0.8156 | -0.8134 |
| 0.0001 | 26.67 | 1000 | 0.9835 | -10.6241 | -14.4122 | 0.7300 | 3.7881 | -258.5334 | -217.0869 | -0.8154 | -0.8130 |
Framework versions
- PEFT 0.10.0
- Transformers 4.39.3
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 1
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for guoyu-zhang/model_hh_usp3_dpo1
Base model
meta-llama/Llama-2-7b-chat-hf