model_shp4_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: 1.7562
- Rewards/chosen: -8.5308
- Rewards/rejected: -8.4695
- Rewards/accuracies: 0.5200
- Rewards/margins: -0.0613
- Logps/rejected: -331.4397
- Logps/chosen: -314.9189
- Logits/rejected: -1.1613
- Logits/chosen: -1.1692
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0922 | 2.67 | 100 | 1.1410 | -4.1724 | -4.0602 | 0.5600 | -0.1122 | -287.3470 | -271.3348 | -0.9400 | -0.9462 |
| 0.0014 | 5.33 | 200 | 1.6279 | -8.0256 | -7.9377 | 0.5400 | -0.0879 | -326.1222 | -309.8669 | -1.2061 | -1.2156 |
| 0.0001 | 8.0 | 300 | 1.6781 | -7.8271 | -7.7492 | 0.4900 | -0.0780 | -324.2366 | -307.8824 | -1.1931 | -1.2019 |
| 0.0001 | 10.67 | 400 | 1.7244 | -8.2046 | -8.1268 | 0.5100 | -0.0778 | -328.0134 | -311.6574 | -1.1773 | -1.1864 |
| 0.0001 | 13.33 | 500 | 1.7449 | -8.3826 | -8.3126 | 0.5100 | -0.0701 | -329.8707 | -313.4376 | -1.1689 | -1.1774 |
| 0.0001 | 16.0 | 600 | 1.7522 | -8.4707 | -8.4001 | 0.5100 | -0.0706 | -330.7461 | -314.3180 | -1.1649 | -1.1729 |
| 0.0001 | 18.67 | 700 | 1.7553 | -8.5177 | -8.4517 | 0.5200 | -0.0659 | -331.2625 | -314.7882 | -1.1626 | -1.1704 |
| 0.0001 | 21.33 | 800 | 1.7608 | -8.5360 | -8.4723 | 0.5200 | -0.0637 | -331.4679 | -314.9713 | -1.1608 | -1.1692 |
| 0.0001 | 24.0 | 900 | 1.7653 | -8.5361 | -8.4664 | 0.5200 | -0.0697 | -331.4087 | -314.9720 | -1.1617 | -1.1693 |
| 0.0001 | 26.67 | 1000 | 1.7562 | -8.5308 | -8.4695 | 0.5200 | -0.0613 | -331.4397 | -314.9189 | -1.1613 | -1.1692 |
Framework versions
- PEFT 0.10.0
- Transformers 4.39.1
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 4
Model tree for guoyu-zhang/model_shp4_dpo1
Base model
meta-llama/Llama-2-7b-chat-hf