model_usp2_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: 2.0083
- Rewards/chosen: -12.0382
- Rewards/rejected: -17.6824
- Rewards/accuracies: 0.6600
- Rewards/margins: 5.6442
- Logps/rejected: -145.7682
- Logps/chosen: -134.7426
- Logits/rejected: -1.0481
- Logits/chosen: -1.0149
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0339 | 2.67 | 100 | 1.1175 | -5.1390 | -8.2341 | 0.6900 | 3.0951 | -126.8717 | -120.9441 | -0.3625 | -0.3504 |
| 0.0325 | 5.33 | 200 | 1.6541 | -3.3759 | -6.0514 | 0.6000 | 2.6755 | -122.5063 | -117.4180 | -0.3455 | -0.3686 |
| 0.0021 | 8.0 | 300 | 2.0449 | -11.1907 | -16.3902 | 0.6500 | 5.1995 | -143.1838 | -133.0475 | -1.0202 | -0.9909 |
| 0.0 | 10.67 | 400 | 2.0215 | -12.0042 | -17.6127 | 0.6600 | 5.6085 | -145.6288 | -134.6745 | -1.0461 | -1.0130 |
| 0.0 | 13.33 | 500 | 2.0139 | -12.0087 | -17.6260 | 0.6700 | 5.6174 | -145.6555 | -134.6835 | -1.0475 | -1.0146 |
| 0.0 | 16.0 | 600 | 2.0087 | -12.0288 | -17.6593 | 0.6700 | 5.6304 | -145.7220 | -134.7238 | -1.0479 | -1.0150 |
| 0.0 | 18.67 | 700 | 2.0128 | -12.0402 | -17.6571 | 0.6700 | 5.6170 | -145.7177 | -134.7465 | -1.0484 | -1.0150 |
| 0.0 | 21.33 | 800 | 2.0087 | -12.0427 | -17.6819 | 0.6600 | 5.6391 | -145.7672 | -134.7516 | -1.0483 | -1.0151 |
| 0.0 | 24.0 | 900 | 2.0120 | -12.0409 | -17.6812 | 0.6700 | 5.6402 | -145.7658 | -134.7480 | -1.0489 | -1.0159 |
| 0.0 | 26.67 | 1000 | 2.0083 | -12.0382 | -17.6824 | 0.6600 | 5.6442 | -145.7682 | -134.7426 | -1.0481 | -1.0149 |
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
- 2
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for guoyu-zhang/model_usp2_dpo5
Base model
meta-llama/Llama-2-7b-chat-hf