model_hh_usp1_dpo5
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: 1.6912
- Rewards/chosen: -8.0278
- Rewards/rejected: -12.3890
- Rewards/accuracies: 0.7100
- Rewards/margins: 4.3613
- Logps/rejected: -138.5718
- Logps/chosen: -128.7695
- Logits/rejected: -0.7832
- Logits/chosen: -0.7072
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1062 | 2.67 | 100 | 0.9731 | -1.3173 | -2.6971 | 0.6900 | 1.3798 | -119.1878 | -115.3486 | -0.0430 | -0.0352 |
| 0.0053 | 5.33 | 200 | 1.4158 | -3.1407 | -7.2838 | 0.6700 | 4.1431 | -128.3614 | -118.9955 | -0.6209 | -0.5447 |
| 0.0359 | 8.0 | 300 | 1.6639 | -5.1494 | -9.1017 | 0.6900 | 3.9524 | -131.9972 | -123.0127 | -0.7498 | -0.6842 |
| 0.0 | 10.67 | 400 | 1.6816 | -7.9965 | -12.3589 | 0.7000 | 4.3624 | -138.5115 | -128.7070 | -0.7830 | -0.7067 |
| 0.0 | 13.33 | 500 | 1.6979 | -8.0167 | -12.3847 | 0.7000 | 4.3680 | -138.5631 | -128.7474 | -0.7829 | -0.7068 |
| 0.0 | 16.0 | 600 | 1.6968 | -8.0295 | -12.3907 | 0.7000 | 4.3611 | -138.5751 | -128.7731 | -0.7829 | -0.7066 |
| 0.0 | 18.67 | 700 | 1.6993 | -8.0304 | -12.3986 | 0.7000 | 4.3682 | -138.5910 | -128.7749 | -0.7833 | -0.7066 |
| 0.0 | 21.33 | 800 | 1.6953 | -8.0311 | -12.4055 | 0.7000 | 4.3743 | -138.6046 | -128.7763 | -0.7835 | -0.7067 |
| 0.0 | 24.0 | 900 | 1.6919 | -8.0283 | -12.4122 | 0.7100 | 4.3839 | -138.6181 | -128.7706 | -0.7836 | -0.7075 |
| 0.0 | 26.67 | 1000 | 1.6912 | -8.0278 | -12.3890 | 0.7100 | 4.3613 | -138.5718 | -128.7695 | -0.7832 | -0.7072 |
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
- 3
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for guoyu-zhang/model_hh_usp1_dpo5
Base model
meta-llama/Llama-2-7b-chat-hf