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: 0.4292
- Rewards/chosen: -1.8869
- Rewards/rejected: -2.7914
- Rewards/accuracies: 0.8242
- Rewards/margins: 0.9045
- Logps/rejected: -612.2493
- Logps/chosen: -524.2042
- Logits/rejected: -0.4436
- Logits/chosen: -0.8025
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5405 | 0.12 | 100 | 0.6086 | -0.8599 | -1.1867 | 0.6953 | 0.3268 | -451.7755 | -421.5048 | -1.6547 | -1.7462 |
| 0.4371 | 0.23 | 200 | 0.5454 | -2.0208 | -2.5842 | 0.7422 | 0.5634 | -591.5291 | -537.5920 | -0.7151 | -0.8867 |
| 0.4348 | 0.35 | 300 | 0.5012 | -2.0998 | -2.8410 | 0.7734 | 0.7413 | -617.2101 | -545.4883 | -0.3499 | -0.5939 |
| 0.3733 | 0.46 | 400 | 0.4721 | -2.1506 | -2.9308 | 0.7773 | 0.7802 | -626.1902 | -550.5717 | -0.2280 | -0.5456 |
| 0.3689 | 0.58 | 500 | 0.4484 | -2.0467 | -2.9485 | 0.7969 | 0.9018 | -627.9595 | -540.1826 | -0.1091 | -0.4774 |
| 0.3829 | 0.69 | 600 | 0.4419 | -2.0265 | -2.9075 | 0.8086 | 0.8810 | -623.8541 | -538.1624 | -0.1412 | -0.5099 |
| 0.3725 | 0.81 | 700 | 0.4329 | -1.9184 | -2.8079 | 0.8242 | 0.8895 | -613.8932 | -527.3496 | -0.3224 | -0.6920 |
| 0.4052 | 0.92 | 800 | 0.4292 | -1.8869 | -2.7914 | 0.8242 | 0.9045 | -612.2493 | -524.2042 | -0.4436 | -0.8025 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2