<!-- 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-wiki_DPO_ds_RM_top_2_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: 0.6572
- Rewards/chosen: -0.1473
- Rewards/rejected: -0.2755
- Rewards/accuracies: 0.6128
- Rewards/margins: 0.1282
- Logps/rejected: -203.3539
- Logps/chosen: -207.2538
- Logits/rejected: 1.1534
- Logits/chosen: 1.1690
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6925 | 0.1 | 19 | 0.6761 | -0.1021 | -0.1593 | 0.5697 | 0.0573 | -202.1919 | -206.8013 | 1.1506 | 1.1664 |
| 0.6754 | 0.21 | 38 | 0.6738 | -0.4156 | -0.5460 | 0.5701 | 0.1303 | -206.0580 | -209.9368 | 1.1257 | 1.1406 |
| 0.6799 | 0.31 | 57 | 0.6666 | -0.0458 | -0.1454 | 0.5932 | 0.0996 | -202.0523 | -206.2388 | 1.1176 | 1.1327 |
| 0.6618 | 0.42 | 76 | 0.6637 | -0.1458 | -0.2745 | 0.5971 | 0.1286 | -203.3434 | -207.2391 | 1.1195 | 1.1333 |
| 0.6706 | 0.52 | 95 | 0.6607 | -0.0386 | -0.1827 | 0.5971 | 0.1440 | -202.4252 | -206.1670 | 1.1334 | 1.1484 |
| 0.668 | 0.63 | 114 | 0.6596 | -0.1615 | -0.2945 | 0.6035 | 0.1330 | -203.5434 | -207.3955 | 1.1500 | 1.1661 |
| 0.6712 | 0.73 | 133 | 0.6597 | -0.1703 | -0.2905 | 0.5979 | 0.1202 | -203.5037 | -207.4840 | 1.1515 | 1.1672 |
| 0.6715 | 0.84 | 152 | 0.6588 | -0.1516 | -0.2745 | 0.6100 | 0.1229 | -203.3436 | -207.2964 | 1.1532 | 1.1691 |
| 0.673 | 0.94 | 171 | 0.6572 | -0.1473 | -0.2755 | 0.6128 | 0.1282 | -203.3539 | -207.2538 | 1.1534 | 1.1690 |
Framework versions
- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
