llama-7b-SFT-qlora-wiki_DPO_ds_RM_random_1024_r_64_alpha_16
This model is a fine-tuned version of dhmeltzer/llama-7b-SFT_ds_wiki65k_1024_r_64_alpha_16_merged on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6801
- Rewards/chosen: -0.1790
- Rewards/rejected: -0.2369
- Rewards/accuracies: 0.5469
- Rewards/margins: 0.0578
- Logps/rejected: -206.1121
- Logps/chosen: -202.9860
- Logits/rejected: 1.1465
- Logits/chosen: 1.1674
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: 0.0002
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6904 | 0.1 | 19 | 0.6904 | -0.3143 | -0.3636 | 0.5458 | 0.0493 | -207.3793 | -204.3384 | 1.1224 | 1.1416 |
| 0.6725 | 0.21 | 38 | 0.6850 | -0.3901 | -0.4540 | 0.5547 | 0.0640 | -208.2836 | -205.0964 | 1.1270 | 1.1469 |
| 0.6818 | 0.31 | 57 | 0.6801 | -0.1790 | -0.2369 | 0.5469 | 0.0578 | -206.1121 | -202.9860 | 1.1465 | 1.1674 |
| 0.6671 | 0.41 | 76 | 0.6863 | -0.2598 | -0.3469 | 0.5580 | 0.0871 | -207.2126 | -203.7936 | 1.1468 | 1.1665 |
| 0.6683 | 0.52 | 95 | 0.6841 | -0.1475 | -0.2325 | 0.5502 | 0.0851 | -206.0687 | -202.6704 | 1.1388 | 1.1590 |
| 0.6626 | 0.62 | 114 | 0.6846 | -0.0836 | -0.1600 | 0.5480 | 0.0764 | -205.3429 | -202.0314 | 1.1263 | 1.1474 |
| 0.6593 | 0.72 | 133 | 0.6864 | -0.1272 | -0.2184 | 0.5625 | 0.0912 | -205.9276 | -202.4675 | 1.1106 | 1.1306 |
| 0.672 | 0.83 | 152 | 0.6857 | -0.1452 | -0.2334 | 0.5592 | 0.0882 | -206.0777 | -202.6477 | 1.1086 | 1.1293 |
| 0.6671 | 0.93 | 171 | 0.6855 | -0.1472 | -0.2350 | 0.5547 | 0.0878 | -206.0934 | -202.6673 | 1.1071 | 1.1270 |
Framework versions
- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3