metadata
license: apache-2.0
base_model: mistralai/Mistral-7B-v0.1
tags:
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-lora
results: []
zephyr-7b-dpo-lora
This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5266
- Rewards/chosen: -0.1470
- Rewards/rejected: -0.8985
- Rewards/accuracies: 0.7460
- Rewards/margins: 0.7516
- Logps/rejected: -228.2694
- Logps/chosen: -266.1304
- Logits/rejected: -1.9412
- Logits/chosen: -2.0659
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: 2
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 8
- total_train_batch_size: 64
- total_eval_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5501 | 1.0 | 968 | 0.5560 | -0.1093 | -0.6400 | 0.7200 | 0.5307 | -225.6847 | -265.7538 | -1.9901 | -2.1115 |
| 0.5412 | 2.0 | 1936 | 0.5318 | -0.1497 | -0.8640 | 0.7420 | 0.7143 | -227.9245 | -266.1583 | -1.9509 | -2.0748 |
| 0.5454 | 3.0 | 2904 | 0.5266 | -0.1470 | -0.8985 | 0.7460 | 0.7516 | -228.2694 | -266.1304 | -1.9412 | -2.0659 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1