- 了解機器學習基礎知識
- 具有大型語言模型（LLMs）的相關經驗
- 熟悉Python和PyTorch
目標受眾
- 數據科學家
- AI工程師
參數高效Fine-Tuning（PEFT）是一系列技術，通過僅修改大型語言模型（LLMs）的一小部分參數，實現高效適應。
這項由講師指導的培訓（線上或線下）面向中級數據科學家和AI工程師，他們希望使用LoRA、Adapter Tuning和Prefix Tuning等方法，以更經濟高效的方式微調大型語言模型。
培訓結束後，參與者將能夠：
- 理解參數高效微調方法的理論。
- 使用Hugging Face PEFT實現LoRA、Adapter Tuning和Prefix Tuning。
- 比較PEFT方法與完整微調的性能和成本權衡。
- 部署和擴展微調後的LLMs，並減少計算和存儲需求。
課程形式
- 互動式講座和討論。
- 大量練習和實踐。
- 在實時實驗室環境中進行動手操作。
課程定制選項
- 如需為本課程定制培訓，請聯繫我們安排。
參數高效微調 (PEFT) 介紹
- 全面微調的動機與限制
- PEFT 概述：目標與優勢
- 在行業中的應用與使用案例
LoRA (低秩適應)
- LoRA 的概念與直覺
- 使用 Hugging Face 和 PyTorch 實現 LoRA
- 實作：使用 LoRA 微調模型
適配器微調
- 適配器模組的工作原理
- 與基於 Transformer 模型的整合
- 實作：將適配器微調應用於 Transformer 模型
前綴微調
- 使用軟提示進行微調
- 與 LoRA 和適配器相比的優勢與限制
- 實作：在 LLM 任務上進行前綴微調
評估與比較 PEFT 方法
- 評估性能與效率的指標
- 訓練速度、記憶體使用與準確性之間的權衡
- 基準測試實驗與結果解釋
部署微調模型
- 保存與載入微調模型
- 基於 PEFT 模型的部署考量
- 整合到應用與管道中
最佳實踐與擴展
- 將 PEFT 與量化與蒸餾結合
- 在低資源與多語言環境中的使用
- 未來方向與活躍研究領域
總結與下一步
United Arab Emirates - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Qatar - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Egypt - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Saudi Arabia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
South Africa - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Brasil - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Canada - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
中国 - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
香港 - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
澳門 - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
台灣 - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
USA - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Österreich - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Schweiz - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Deutschland - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Czech Republic - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Denmark - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Estonia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Finland - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Greece - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Magyarország - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Ireland - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Luxembourg - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Latvia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
España - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Italia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Lithuania - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Nederland - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Norway - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Portugal - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
România - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Sverige - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Türkiye - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Malta - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Belgique - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
France - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
日本 - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Australia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Malaysia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
New Zealand - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Philippines - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Singapore - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Thailand - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Vietnam - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
India - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Argentina - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Chile - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Costa Rica - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Ecuador - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Guatemala - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Colombia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
México - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Panama - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Peru - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Uruguay - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Venezuela - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Polska - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
United Kingdom - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
South Korea - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Pakistan - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Sri Lanka - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Bulgaria - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Bolivia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Indonesia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Kazakhstan - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Moldova - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Morocco - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Tunisia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Kuwait - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Oman - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Slovakia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Kenya - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Nigeria - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Botswana - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Slovenia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Croatia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Serbia - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Bhutan - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Nepal - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs
Uzbekistan - Parameter-Efficient Fine-Tuning (PEFT) Techniques for LLMs