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: 0.0544
- Rewards/chosen: -2.1973
- Rewards/rejected: -2.9243
- Rewards/accuracies: 0.7070
- Rewards/margins: 0.7270
- Logps/rejected: -549.7877
- Logps/chosen: -476.7722
- Logits/rejected: -1.9407
- Logits/chosen: -1.9849
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: 3
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 4
- total_train_batch_size: 256
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0846 | 0.23 | 100 | 0.0846 | -1.4642 | -1.8940 | 0.6484 | 0.4298 | -446.7535 | -403.4620 | -2.3302 | -2.3522 |
| 0.0477 | 0.45 | 200 | 0.0672 | -1.7958 | -2.4017 | 0.7148 | 0.6059 | -497.5217 | -436.6205 | -2.1284 | -2.1617 |
| 0.046 | 0.68 | 300 | 0.0552 | -2.1484 | -2.8722 | 0.7148 | 0.7238 | -544.5698 | -471.8781 | -1.9484 | -1.9914 |
| 0.0439 | 0.91 | 400 | 0.0544 | -2.1973 | -2.9243 | 0.7070 | 0.7270 | -549.7877 | -476.7722 | -1.9407 | -1.9849 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1