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: 0.4905
- Rewards/chosen: -0.7983
- Rewards/rejected: -1.8319
- Rewards/accuracies: 0.7539
- Rewards/margins: 1.0337
- Logps/rejected: -470.9522
- Logps/chosen: -361.4586
- Logits/rejected: -0.7601
- Logits/chosen: -0.9126
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5759 | 0.21 | 100 | 0.5723 | -0.5418 | -1.0793 | 0.7148 | 0.5375 | -395.6910 | -335.8161 | -1.8988 | -1.9151 |
| 0.5391 | 0.42 | 200 | 0.5138 | -0.3953 | -1.1724 | 0.7461 | 0.7771 | -405.0056 | -321.1682 | -1.5774 | -1.6570 |
| 0.4788 | 0.63 | 300 | 0.4939 | -0.8609 | -1.8779 | 0.7578 | 1.0170 | -475.5532 | -367.7277 | -0.8826 | -0.9864 |
| 0.4937 | 0.84 | 400 | 0.4905 | -0.7983 | -1.8319 | 0.7539 | 1.0337 | -470.9522 | -361.4586 | -0.7601 | -0.9126 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2