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: 0.0427
- Rewards/chosen: -1.1712
- Rewards/rejected: -2.0556
- Rewards/accuracies: 0.7266
- Rewards/margins: 0.8844
- Logps/rejected: -517.1834
- Logps/chosen: -420.7032
- Logits/rejected: -0.0713
- Logits/chosen: -0.0801
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: 3e-06
- train_batch_size: 4
- eval_batch_size: 8
- seed: 2
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 4
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0642 | 0.21 | 100 | 0.0487 | -1.3984 | -2.4508 | 0.7188 | 1.0524 | -556.7064 | -443.4212 | 0.4579 | 0.4301 |
| 0.047 | 0.42 | 200 | 0.0461 | -1.1146 | -1.8461 | 0.7422 | 0.7314 | -496.2327 | -415.0494 | 0.1274 | 0.1123 |
| 0.0401 | 0.63 | 300 | 0.0408 | -1.2816 | -2.1650 | 0.7148 | 0.8834 | -528.1252 | -431.7439 | -0.1765 | -0.1928 |
| 0.0435 | 0.84 | 400 | 0.0427 | -1.1712 | -2.0556 | 0.7266 | 0.8844 | -517.1834 | -420.7032 | -0.0713 | -0.0801 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1