# PEFT

**Repository Path**: xpnb/peft

## Basic Information

- **Project Name**: PEFT
- **Description**: PEFT论文阅读
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2023-10-24
- **Last Updated**: 2023-12-04

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# PEFT(Parameter-Efficient Fine-tune)论文阅读与汇报

我们按照论文“[LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models](http://arxiv.org/abs/2304.01933)”的分类法将PEFT论文分为：
- Prompt-based learning.
- Reparametrization-based method.
- Series Adapter.
- Parallel Adapter.

每位同学选一个类别的两篇论文进行阅读，我做最后的汇总、梳理、补充和presentation。

**注意：** 不要更改论文阅读笔记模板，模板自己copy一份并用 `方法名+姓名拼音首字母` 命名再修改，如：

LoRA_lxp.md

## Prompt-based learning.
### TOREAD List：
- [x] Prompt Tuning: [The Power of Scale for Parameter-Efficient Prompt Tuning](https://arxiv.org/pdf/2104.08691.pdf) 
- [x] Prefix Tuning: [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://aclanthology.org/2021.acl-long.353/),[P-tuning: GPT Understands, Too](https://arxiv.org/abs/2103.10385),  [P-Tuning v2: Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Tasks](https://arxiv.org/pdf/2110.07602.pdf),

## Reparametrization-based method.
### TOREAD List：
- [x] LoRA: [LORA: LOW-RANK ADAPTATION OF LARGE LANGUAGE MODELS](https://arxiv.org/pdf/2106.09685.pdf)
- [x] QLoRA: [QLoRA: Efficient finetuning of quantized LLMs](https://arxiv.org/abs/2305.14314)

## Series Adapter.
### TOREAD List：
- [x] Adapter: [Parameter-Efficient Transfer Learning for NLP](https://arxiv.org/pdf/1902.00751.pdf)
- [x] AdapterFusion: [AdapterFusion: Non-Destructive Task Composition for Transfer Learning](https://arxiv.org/abs/2005.00247)

## Parallel Adapter.
### TOREAD List：
- [x] Parallel: [TOWARDS A UNIFIED VIEW OF PARAMETER-EFFICIENT TRANSFER LEARNING](https://arxiv.org/pdf/2110.04366.pdf)
- [x] Lst: [Lst: Ladder side-tuning for parameter and memory efficient transfer learning](https://proceedings.neurips.cc/paper_files/paper/2022/file/54801e196796134a2b0ae5e8adef502f-Paper-Conference.pdf)

## 其他
- LoRA带来的安全性问题：[LoRA Fine-tuning Efficiently Undoes Safety Training in Llama 2-Chat 70B](https://arxiv.org/pdf/2310.20624v1.pdf)
- 如何有效使用PEFT？[Empirical Analysis of the Strengths and Weaknesses of PEFT Techniques for LLMs | PDF (arxiv.org)](https://arxiv.org/pdf/2304.14999.pdf)
- 如何使用PEFT？[Scaling Down to Scale Up: A Guide to Parameter-Efficient Fine-Tuning | PDF (arxiv.org)](https://arxiv.org/pdf/2303.15647.pdf)
- PEFT best practice: [Parameter-Efficient Fine-Tuning (PEFT) of LLMs: A Practical Guide](https://markovate.com/blog/parameter-efficient-fine-tuning-peft-of-llms-a-practical-guide/)

## 参考资料
### 知乎：
- [让天下没有难Tuning的大模型-PEFT技术简介](https://zhuanlan.zhihu.com/p/618894319)
- [Hugging Face模型高效微调工具peft详解](https://zhuanlan.zhihu.com/p/646611666)
- [大模型参数高效微调(PEFT)](https://zhuanlan.zhihu.com/p/621700272)
- [高效调参-PEFT库简介及使用](https://zhuanlan.zhihu.com/p/624393922)
- [大模型的领域适配 —— Parameter-Efficient Fine-Tuning (PEFT)](https://zhuanlan.zhihu.com/p/636326003)
- [QLoRA](https://www.zhihu.com/question/593383416/answer/3099434564)
- [LORA微调系列](https://www.zhihu.com/people/li-ge-jiu-shi-666/posts)
### paper：
- [Parameter-efficient fine-tuning of large-scale pre-trained language models](https://www.nature.com/articles/s42256-023-00626-4)