zephyr-7b-dpo-lora
This model is a fine-tuned version of decapoda-research/Adrastea-7b-v1.0 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6086
- Rewards/chosen: 0.1339
- Rewards/rejected: -0.0905
- Rewards/accuracies: 0.6976
- Rewards/margins: 0.2244
- Logps/rejected: -260.8578
- Logps/chosen: -297.0598
- Logits/rejected: -3.6473
- Logits/chosen: -3.5958
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: 420
- distributed_type: multi-GPU
- num_devices: 3
- gradient_accumulation_steps: 32
- total_train_batch_size: 192
- total_eval_batch_size: 12
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6101 | 1.0 | 322 | 0.6086 | 0.1339 | -0.0905 | 0.6976 | 0.2244 | -260.8578 | -297.0598 | -3.6473 | -3.5958 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1
- Downloads last month
- 5