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: 1.1360
- Rewards/chosen: -0.1815
- Rewards/rejected: 0.2981
- Rewards/accuracies: 0.4102
- Rewards/margins: -0.4796
- Logps/rejected: -515.4065
- Logps/chosen: -392.2896
- Logits/rejected: -4.7596
- Logits/chosen: -4.5397
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-08
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2984 | 0.26 | 100 | 0.9403 | -0.0277 | 0.3050 | 0.4062 | -0.3328 | -515.3372 | -390.7516 | -4.7620 | -4.5507 |
| 0.2338 | 0.51 | 200 | 1.0997 | -0.0605 | 0.4142 | 0.3867 | -0.4747 | -514.2458 | -391.0792 | -4.7584 | -4.5386 |
| 0.2158 | 0.77 | 300 | 1.1360 | -0.1815 | 0.2981 | 0.4102 | -0.4796 | -515.4065 | -392.2896 | -4.7596 | -4.5397 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2