大语言模型微调入门：从 ChatGPT 训练原理到本地 QLoRA 实作
本文是 AI 系列的第三篇。如果你还不了解 AI 的基本概念与 Transformer 发展脉络，建议先阅读《从智慧到人工智能：AI 基本概念与发展史入门》与《机器学习与 MNIST 手写数字识别入门》。 这篇文章会从 ChatGPT 等大语言模型「怎么被训练出来」讲起，解释为什么预训练 alone 不够、为什么需要 SFT 与 RLHF 做对齐，最后带你用...
全部文章
技术笔记、教程与思考。
筛选
清除筛选分类
标签
全部标签#Agent-Gateway#AGENTS.md#AI#AI-Agent#AI-History#Aircrack-ng#Alignment#Artificial-Intelligence#ASIC#Autonomous-Agent#Billboard#Breaking-Changes#Campus Life#CLI-Tool#CNN#Company#Computer#Computer-Network#CPU#Cryptography#Cursor#Datacenter#datasheet#Deep-Learning#Dell#Dependency#Developer-Tools#Digital Design#DRAM#EDA#Education#Fans Speed#Fine-Tuning#FPGA#Front-End#FSM#Game#Game-Development#GDG#Gemini#Go#Google-Gemini#Graphics#Hard Link#Hashcat#History#Intel#Intern#JavaScript#Job#JSX#Legacy-Code#LibreLane#Linkou#Linux#LLM#LoRA#Machine-Learning#macOS#Memory Controller#Migration#MNIST#MPW#NAT#NATBypass#Neural-Network#node_modules#Node.js#Normal-Mapping#npm#NTNU#OpenClaw#OpenCode#OpenROAD#Operating System#Operations#PDK#Penetration-Testing#pnpm#Policy Speech#Port-Forwarding#PowerEdge#Prompt#Prompt-Engineering#PUE#Python#PyTorch#QLoRA#Qwen#React#Resign#Reverse-Tunnel#RLHF#RTL#RTL-to-GDSII#SDRAM#Server#SFT#SkyWater130#Slido#SRAM#Student Election#Student Government#Synthesis#SystemVerilog#Taipei Higher School#Taiwan#Tapeout#TCP#Timing#Tiny Tapeout#Transformer#Trick#TypeScript#Typography#University#User Interface(UI)#UX(User Experience)#Verilog#Vibe-Coding#Virtual DOM#Web#Web Front-End#WEP#WiFi#Windows#Wireless-Security#Work#WPA#WPA2#WPA3#Yosys