model_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: 1.2527
- Rewards/chosen: -10.7037
- Rewards/rejected: -13.2986
- Rewards/accuracies: 0.7000
- Rewards/margins: 2.5949
- Logps/rejected: -242.7872
- Logps/chosen: -215.0880
- Logits/rejected: -0.8066
- Logits/chosen: -0.8008
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1152 | 2.67 | 100 | 0.9230 | -6.0994 | -7.5730 | 0.6100 | 1.4737 | -185.5316 | -169.0449 | -0.9595 | -0.9289 |
| 0.0007 | 5.33 | 200 | 1.2625 | -10.4507 | -12.9431 | 0.7100 | 2.4924 | -239.2319 | -212.5582 | -0.8924 | -0.8878 |
| 0.0002 | 8.0 | 300 | 1.2065 | -10.0237 | -12.4963 | 0.7000 | 2.4725 | -234.7639 | -208.2885 | -0.8455 | -0.8351 |
| 0.0001 | 10.67 | 400 | 1.2314 | -10.3811 | -12.9055 | 0.7100 | 2.5245 | -238.8566 | -211.8620 | -0.8259 | -0.8181 |
| 0.0001 | 13.33 | 500 | 1.2449 | -10.5483 | -13.1112 | 0.7100 | 2.5629 | -240.9136 | -213.5344 | -0.8155 | -0.8090 |
| 0.0001 | 16.0 | 600 | 1.2475 | -10.6353 | -13.2168 | 0.7100 | 2.5815 | -241.9690 | -214.4042 | -0.8099 | -0.8041 |
| 0.0001 | 18.67 | 700 | 1.2504 | -10.6796 | -13.2671 | 0.7000 | 2.5875 | -242.4725 | -214.8474 | -0.8075 | -0.8022 |
| 0.0001 | 21.33 | 800 | 1.2562 | -10.7029 | -13.2944 | 0.7000 | 2.5915 | -242.7449 | -215.0800 | -0.8065 | -0.8014 |
| 0.0001 | 24.0 | 900 | 1.2542 | -10.7066 | -13.2945 | 0.7000 | 2.5879 | -242.7467 | -215.1174 | -0.8061 | -0.8009 |
| 0.0001 | 26.67 | 1000 | 1.2527 | -10.7037 | -13.2986 | 0.7000 | 2.5949 | -242.7872 | -215.0880 | -0.8066 | -0.8008 |
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_usp3_dpo1
Base model
meta-llama/Llama-2-7b-chat-hf