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: 0.0270
- Rewards/chosen: -1.1958
- Rewards/rejected: -1.8757
- Rewards/accuracies: 0.7266
- Rewards/margins: 0.6799
- Logps/rejected: -444.9223
- Logps/chosen: -376.6192
- Logits/rejected: -2.4111
- Logits/chosen: -2.4251
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0517 | 0.21 | 100 | 0.0427 | -0.6331 | -1.0207 | 0.7031 | 0.3876 | -359.4206 | -320.3455 | -2.7320 | -2.7523 |
| 0.0305 | 0.42 | 200 | 0.0297 | -1.0902 | -1.6743 | 0.7227 | 0.5842 | -424.7881 | -366.0565 | -2.5649 | -2.5797 |
| 0.0258 | 0.63 | 300 | 0.0274 | -1.2031 | -1.8719 | 0.7188 | 0.6688 | -444.5428 | -377.3524 | -2.4247 | -2.4384 |
| 0.0234 | 0.84 | 400 | 0.0270 | -1.1958 | -1.8757 | 0.7266 | 0.6799 | -444.9223 | -376.6192 | -2.4111 | -2.4251 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1