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: 0.2511
- Rewards/chosen: 14.1512
- Rewards/rejected: -27.0299
- Rewards/accuracies: 0.9297
- Rewards/margins: 41.1811
- Logps/rejected: -120.2706
- Logps/chosen: -123.6211
- Logits/rejected: -1.8742
- Logits/chosen: -1.8698
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1812 | 0.21 | 100 | 0.1474 | 12.7707 | -20.3277 | 0.9180 | 33.0984 | -113.5685 | -125.0015 | -1.7088 | -1.7301 |
| 0.2958 | 0.42 | 200 | 0.2224 | 15.4746 | -23.1680 | 0.9258 | 38.6426 | -116.4087 | -122.2977 | -1.8350 | -1.8384 |
| 0.3034 | 0.63 | 300 | 0.2672 | 14.1732 | -27.0300 | 0.9258 | 41.2032 | -120.2707 | -123.5991 | -1.8525 | -1.8496 |
| 0.3576 | 0.84 | 400 | 0.2511 | 14.1512 | -27.0299 | 0.9297 | 41.1811 | -120.2706 | -123.6211 | -1.8742 | -1.8698 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2