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: 6.9230
- Rewards/chosen: -4.5175
- Rewards/rejected: 0.4288
- Rewards/accuracies: 0.3164
- Rewards/margins: -4.9464
- Logps/rejected: -517.5300
- Logps/chosen: -399.5095
- Logits/rejected: -4.8908
- Logits/chosen: -4.6604
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5909 | 0.2635 | 100 | 6.6534 | -3.0090 | 2.8904 | 0.2773 | -5.8994 | -512.6068 | -396.4924 | -4.8508 | -4.6121 |
| 0.7239 | 0.5270 | 200 | 8.0720 | -2.8191 | 3.5065 | 0.2734 | -6.3256 | -511.3747 | -396.1127 | -4.9896 | -4.7715 |
| 0.5556 | 0.7905 | 300 | 6.9230 | -4.5175 | 0.4288 | 0.3164 | -4.9464 | -517.5300 | -399.5095 | -4.8908 | -4.6604 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.19.1