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: 0.2143
- Rewards/chosen: -0.8956
- Rewards/rejected: -1.5167
- Rewards/accuracies: 0.7031
- Rewards/margins: 0.6212
- Logps/rejected: -409.0278
- Logps/chosen: -346.5983
- Logits/rejected: -2.4275
- Logits/chosen: -2.4425
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: 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2667 | 0.21 | 100 | 0.2670 | -0.4530 | -0.7921 | 0.6797 | 0.3391 | -336.5636 | -302.3352 | -2.7593 | -2.7741 |
| 0.2068 | 0.42 | 200 | 0.2087 | -0.8343 | -1.3671 | 0.6836 | 0.5328 | -394.0588 | -340.4660 | -2.5512 | -2.5673 |
| 0.2095 | 0.63 | 300 | 0.2233 | -0.8384 | -1.4377 | 0.7109 | 0.5993 | -401.1194 | -340.8771 | -2.4645 | -2.4791 |
| 0.204 | 0.84 | 400 | 0.2143 | -0.8956 | -1.5167 | 0.7031 | 0.6212 | -409.0278 | -346.5983 | -2.4275 | -2.4425 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1