model_hh_usp4_200
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: 2.2047
- Rewards/chosen: -10.4203
- Rewards/rejected: -11.3883
- Rewards/accuracies: 0.6100
- Rewards/margins: 0.9680
- Logps/rejected: -126.6996
- Logps/chosen: -122.1098
- Logits/rejected: -0.0920
- Logits/chosen: -0.0765
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0 | 8.0 | 100 | 2.1928 | -10.7974 | -11.7956 | 0.6200 | 0.9982 | -127.1522 | -122.5288 | -0.0977 | -0.0806 |
| 0.0 | 16.0 | 200 | 2.2073 | -10.6145 | -11.6045 | 0.6200 | 0.9900 | -126.9398 | -122.3256 | -0.0957 | -0.0785 |
| 0.0 | 24.0 | 300 | 2.1834 | -10.5576 | -11.5387 | 0.6100 | 0.9810 | -126.8667 | -122.2624 | -0.0926 | -0.0766 |
| 0.0 | 32.0 | 400 | 2.2206 | -10.5218 | -11.4563 | 0.6000 | 0.9345 | -126.7752 | -122.2226 | -0.0924 | -0.0768 |
| 0.0 | 40.0 | 500 | 2.1989 | -10.4576 | -11.4408 | 0.6100 | 0.9832 | -126.7580 | -122.1513 | -0.0920 | -0.0763 |
| 0.0 | 48.0 | 600 | 2.1897 | -10.4344 | -11.3970 | 0.6000 | 0.9626 | -126.7093 | -122.1255 | -0.0915 | -0.0758 |
| 0.0 | 56.0 | 700 | 2.1723 | -10.3994 | -11.3863 | 0.6000 | 0.9869 | -126.6974 | -122.0866 | -0.0916 | -0.0760 |
| 0.0 | 64.0 | 800 | 2.1910 | -10.4312 | -11.3832 | 0.6100 | 0.9520 | -126.6939 | -122.1220 | -0.0918 | -0.0760 |
| 0.0 | 72.0 | 900 | 2.1762 | -10.4083 | -11.3782 | 0.6100 | 0.9699 | -126.6885 | -122.0965 | -0.0916 | -0.0762 |
| 0.0 | 80.0 | 1000 | 2.2047 | -10.4203 | -11.3883 | 0.6100 | 0.9680 | -126.6996 | -122.1098 | -0.0920 | -0.0765 |
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
- 0
Model tree for guoyu-zhang/model_hh_usp4_200
Base model
meta-llama/Llama-2-7b-chat-hf