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: 0.0660
- Rewards/chosen: -2.5606
- Rewards/rejected: -2.9549
- Rewards/accuracies: 0.625
- Rewards/margins: 0.3944
- Logps/rejected: -552.8470
- Logps/chosen: -513.0960
- Logits/rejected: -2.2459
- Logits/chosen: -2.2708
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: 4
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0437 | 0.25 | 100 | 0.0824 | -2.2538 | -2.4741 | 0.5859 | 0.2203 | -504.7590 | -482.4154 | -2.3143 | -2.3260 |
| 0.0258 | 0.49 | 200 | 0.0581 | -2.8677 | -3.2192 | 0.5977 | 0.3515 | -579.2755 | -543.8072 | -2.1155 | -2.1394 |
| 0.0402 | 0.74 | 300 | 0.0837 | -2.0997 | -2.5006 | 0.6289 | 0.4009 | -507.4115 | -467.0057 | -2.2751 | -2.2980 |
| 0.0288 | 0.99 | 400 | 0.0660 | -2.5606 | -2.9549 | 0.625 | 0.3944 | -552.8470 | -513.0960 | -2.2459 | -2.2708 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1