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: 239.1993
- Rewards/chosen: 0.3530
- Rewards/rejected: 0.3216
- Rewards/accuracies: 0.5625
- Rewards/margins: 0.0314
- Logps/rejected: -225.1900
- Logps/chosen: -221.7349
- Logits/rejected: -2.6730
- Logits/chosen: -2.6975
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 256.7688 | 0.21 | 100 | 245.7276 | 0.2947 | 0.2688 | 0.5742 | 0.0259 | -230.4730 | -227.5649 | -2.7670 | -2.7863 |
| 239.4804 | 0.42 | 200 | 241.5078 | 0.3319 | 0.3036 | 0.5742 | 0.0283 | -226.9933 | -223.8462 | -2.7082 | -2.7290 |
| 240.5041 | 0.63 | 300 | 239.8053 | 0.3474 | 0.3163 | 0.5625 | 0.0310 | -225.7186 | -222.3044 | -2.6807 | -2.7042 |
| 236.8453 | 0.84 | 400 | 239.1993 | 0.3530 | 0.3216 | 0.5625 | 0.0314 | -225.1900 | -221.7349 | -2.6730 | -2.6975 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1