Krahasoni metodat
Shqyrtoni metodat e zgjedhura krah për krah; rreshtat që ndryshojnë janë të theksuar.
| Optimizimi i Drejtpërdrejtë i Preferencave× | QLoRA× | |
|---|---|---|
| Fusha | Mësimi i thellë | Mësimi i thellë |
| Familja | Machine learning | Machine learning |
| Viti i origjinës | 2023 | 2023 |
| Krijuesi≠ | Rafael Rafailov | Tim Dettmers |
| Lloji | Training methodology | Training methodology |
| Burimi themelues≠ | Rafailov, R., Sharma, A., Mitchell, E., Manning, C. D., Ermon, S., & Finn, C. (2023). Direct preference optimization: Your language model is secretly a reward model. arXiv preprint arXiv:2305.18290. link ↗ | Dettmers, T., Pagnoni, A., Holtzman, A., & Contrastive, L. (2023). QLoRA: Efficient finetuning of quantized LLMs. arXiv preprint arXiv:2305.14314. link ↗ |
| Emërtime të tjera | DPO, Direct preference | QLoRA, Quantized LoRA |
| Të lidhura | 4 | 4 |
| Përmbledhja≠ | Direct Preference Optimization (DPO) is a training method introduced by Rafailov et al. in 2023 that aligns language models with human preferences without requiring an explicit reward model. By directly optimizing for preference pairs (better response vs worse response), DPO simplifies the training pipeline compared to reinforcement learning from human feedback (RLHF). | QLoRA is an efficient fine-tuning method introduced by Dettmers et al. in 2023 that enables fine-tuning large language models using quantization and low-rank adaptation. By combining 4-bit quantization with LoRA, QLoRA reduces memory requirements by 75%, enabling fine-tuning of 65B-parameter models on single GPUs. |
| ScholarGateSeti i të dhënave ↗ |
|
|