metadata
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-full
results: []
zephyr-7b-dpo-full
This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1390
- Rewards/chosen: 3.4895
- Rewards/rejected: -9.2522
- Rewards/accuracies: 0.9297
- Rewards/margins: 12.7417
- Logps/rejected: -139.5015
- Logps/chosen: -120.3246
- Logits/rejected: -1.8106
- Logits/chosen: -1.8098
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: 5e-07
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1523 | 0.21 | 100 | 0.1399 | 2.5441 | -8.9516 | 0.9375 | 11.4956 | -137.9985 | -125.0519 | -1.8014 | -1.8101 |
| 0.176 | 0.42 | 200 | 0.1358 | 3.3974 | -8.7531 | 0.9375 | 12.1505 | -137.0064 | -120.7853 | -1.8762 | -1.8764 |
| 0.1509 | 0.63 | 300 | 0.1403 | 3.3534 | -9.3163 | 0.9336 | 12.6696 | -139.8221 | -121.0054 | -1.7873 | -1.7875 |
| 0.2009 | 0.84 | 400 | 0.1390 | 3.4895 | -9.2522 | 0.9297 | 12.7417 | -139.5015 | -120.3246 | -1.8106 | -1.8098 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2