metadata
license: mit
base_model: HuggingFaceH4/mistral-7b-sft-beta
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-full
results: []
zephyr-7b-dpo-full
This model is a fine-tuned version of HuggingFaceH4/mistral-7b-sft-beta on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0465
- Rewards/chosen: -2.6400
- Rewards/rejected: -3.4900
- Rewards/accuracies: 0.7227
- Rewards/margins: 0.8499
- Logps/rejected: -606.3505
- Logps/chosen: -521.0439
- Logits/rejected: -1.9091
- Logits/chosen: -1.9501
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: 5
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1485 | 0.11 | 100 | 0.1803 | -0.5621 | -0.7737 | 0.6406 | 0.2117 | -334.7263 | -313.2471 | -2.4998 | -2.5133 |
| 0.0592 | 0.23 | 200 | 0.0662 | -1.7402 | -2.3280 | 0.6797 | 0.5878 | -490.1518 | -431.0574 | -2.2396 | -2.2729 |
| 0.0394 | 0.34 | 300 | 0.0494 | -2.3707 | -2.9767 | 0.6953 | 0.6061 | -555.0248 | -494.1047 | -2.1101 | -2.1389 |
| 0.0401 | 0.45 | 400 | 0.0523 | -2.4275 | -3.1076 | 0.7031 | 0.6801 | -568.1116 | -499.7916 | -2.0429 | -2.0799 |
| 0.0335 | 0.57 | 500 | 0.0461 | -2.4063 | -3.2276 | 0.7148 | 0.8213 | -580.1129 | -497.6727 | -2.0057 | -2.0456 |
| 0.0273 | 0.68 | 600 | 0.0409 | -2.8465 | -3.7152 | 0.7070 | 0.8687 | -628.8741 | -541.6862 | -1.9162 | -1.9558 |
| 0.0377 | 0.79 | 700 | 0.0496 | -2.5317 | -3.3682 | 0.7227 | 0.8365 | -594.1712 | -510.2102 | -1.9274 | -1.9673 |
| 0.0352 | 0.91 | 800 | 0.0465 | -2.6400 | -3.4900 | 0.7227 | 0.8499 | -606.3505 | -521.0439 | -1.9091 | -1.9501 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1