ResLoRA:低秩自适应中的恒等残差映射
ResLoRA: Identity Residual Mapping in Low-Rank Adaption
AI总结:
针对 LoRA 因计算路径长而难以快速有效更新权重的问题,本文提出 ResLoRA,在训练时引入残差路径、推理时合并消除,以无额外参数和成本实现更快更优微调。
摘要:
作为最流行的参数高效微调(PEFT)方法之一,低秩自适应(LoRA)通常被用于微调大型语言模型(LLM)。然而,由于原始模型中的计算路径较长,有效且迅速地更新 LoRA 块的权重具有挑战性。为解决这一问题,我们提出 ResLoRA,这是 LoRA 的一个改进框架。通过在训练期间添加残差路径,并在推理期间使用合并方法消除这些额外路径,与 LoRA 相比,我们的方法能够在更少的训练步骤内取得更好的结果,且不会产生任何额外的可训练参数或推理成本。在 NLG、NLU 和文本到图像任务上的实验证明了我们方法的有效性。据我们所知,ResLoRA 是首个将残差路径与 LoRA 相结合的工作。我们方法的代码可在 https://github.com/microsoft/LMOps/tree/main/reslora 获取。
英文摘要:
As one of the most popular parameter-efficient fine-tuning (PEFT) methods, low-rank adaptation (LoRA) is commonly applied to fine-tune large language models (LLMs). However, updating the weights of LoRA blocks effectively and expeditiously is challenging due to the long calculation path in the original model. To address this, we propose ResLoRA, an improved framework of LoRA. By adding residual paths during training and using merging approaches to eliminate these extra paths during inference, our method can achieve better results in fewer training steps without any extra trainable parameters or inference cost compared to LoRA. The experiments on NLG, NLU, and text-to-image tasks demonstrate the effectiveness of our method. To the best of our knowledge, ResLoRA is the first work that combines the residual path with LoRA. The code of our method is available at https://github.com/microsoft/LMOps/tree/main/reslora .
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。