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: 0.0945
- Rewards/chosen: -1.3600
- Rewards/rejected: -2.1836
- Rewards/accuracies: 0.7656
- Rewards/margins: 0.8237
- Logps/rejected: -475.7151
- Logps/chosen: -393.0347
- Logits/rejected: -2.3019
- Logits/chosen: -2.3254
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1643 | 0.21 | 100 | 0.1558 | -0.4076 | -0.7972 | 0.7461 | 0.3896 | -337.0709 | -297.7996 | -2.7691 | -2.7902 |
| 0.1003 | 0.42 | 200 | 0.0997 | -1.2712 | -1.9340 | 0.7031 | 0.6629 | -450.7552 | -384.1553 | -2.5137 | -2.5340 |
| 0.0953 | 0.63 | 300 | 0.1024 | -1.2036 | -1.9243 | 0.7539 | 0.7207 | -449.7823 | -377.3981 | -2.3837 | -2.4030 |
| 0.0811 | 0.84 | 400 | 0.0945 | -1.3600 | -2.1836 | 0.7656 | 0.8237 | -475.7151 | -393.0347 | -2.3019 | -2.3254 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1