Views
No views yet
DPO Trainer
TRL supports the DPO Trainer for training language models from preference data, as described in the paper Direct Preference Optimization: Your Language Model is Secretly a Reward Model by Rafailov et al., 2023. | Metric | Value |
|---|---|
| Avg. | 77.91 |
| AI2 Reasoning Challenge (25-Shot) | 74.06 |
| HellaSwag (10-Shot) | 86.74 |
| MMLU (5-Shot) | 76.65 |
| TruthfulQA (0-shot) | 72.24 |
| Winogrande (5-shot) | 83.35 |
| GSM8k (5-shot) | 74.45 |