Machine learning
LoRA 和 PEFT
LoRA（低秩适应，Low-Rank Adaptation），由 Hu 等人于 2022 年提出，以及更广泛的参数高效微调（PEFT，parameter-efficient fine-tuning）方法家族，通过仅训练少量额外参数而非模型中的所有权重，将大型预训练语言模型适应新任务。这使得微调所需的 GPU 内存和计算量大大减少，同时原始模型基本保持不变。
阅读完整方法
仅限会员
登录使用免费账户登录即可阅读本节。
Method map
The neighbourhood of related methods — select a node to explore.
来源
- Hu, E. J. et al. (2022). LoRA: Low-Rank Adaptation of Large Language Models. ICLR. link ↗
- Lester, B. et al. (2021). The Power of Scale for Parameter-Efficient Prompt Tuning. EMNLP. DOI: 10.18653/v1/2021.emnlp-main.243 ↗
如何引用本页
ScholarGate. (2026, June 1). Low-Rank Adaptation and Parameter-Efficient Fine-Tuning. ScholarGate. https://scholargate.app/zh/deep-learning/lora-peft
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- 生成对抗网络深度学习↔ compare
- 随机森林机器学习↔ compare
- 变分自编码器深度学习↔ compare
- Vision Transformer深度学习↔ compare
- XGBoost机器学习↔ compare