<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
llama-7b-SFT-qlora-eli5_DPO_ds_RM_contrast_1024_r_64_alpha_16
This model is a fine-tuned version of dhmeltzer/llama-7b-SFT_ds_eli5_1024_r_64_alpha_16_merged on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6204
- Rewards/chosen: -0.0937
- Rewards/rejected: -0.3490
- Rewards/accuracies: 0.6641
- Rewards/margins: 0.2553
- Logps/rejected: -205.9560
- Logps/chosen: -211.6011
- Logits/rejected: 1.1663
- Logits/chosen: 1.1890
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6908 | 0.1 | 19 | 0.6536 | -0.2975 | -0.4466 | 0.6060 | 0.1491 | -206.9322 | -213.6386 | 1.1767 | 1.1980 |
| 0.6613 | 0.21 | 38 | 0.6391 | -0.1759 | -0.3858 | 0.6172 | 0.2099 | -206.3239 | -212.4229 | 1.1695 | 1.1930 |
| 0.6667 | 0.31 | 57 | 0.6297 | -0.0287 | -0.2656 | 0.6440 | 0.2369 | -205.1224 | -210.9511 | 1.1612 | 1.1863 |
| 0.6532 | 0.42 | 76 | 0.6271 | -0.0915 | -0.3376 | 0.6172 | 0.2461 | -205.8420 | -211.5791 | 1.1395 | 1.1612 |
| 0.6546 | 0.52 | 95 | 0.6235 | -0.0575 | -0.2906 | 0.6362 | 0.2331 | -205.3723 | -211.2390 | 1.1551 | 1.1781 |
| 0.6528 | 0.62 | 114 | 0.6231 | -0.0939 | -0.3382 | 0.6562 | 0.2443 | -205.8482 | -211.6033 | 1.1702 | 1.1932 |
| 0.646 | 0.73 | 133 | 0.6204 | -0.1525 | -0.4204 | 0.6518 | 0.2678 | -206.6696 | -212.1891 | 1.1664 | 1.1886 |
| 0.6524 | 0.83 | 152 | 0.6208 | -0.1083 | -0.3660 | 0.6607 | 0.2577 | -206.1257 | -211.7465 | 1.1548 | 1.1765 |
| 0.6335 | 0.94 | 171 | 0.6204 | -0.0937 | -0.3490 | 0.6641 | 0.2553 | -205.9560 | -211.6011 | 1.1663 | 1.1890 |
Framework versions
- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
