【推荐内容】合集 · jingyaogong/minimind#504

2025-10-23T12:19:28.000Z

* MiniMind的源码解读，包含tokenizer、RoPE、MoE、KV Cache、pretraining、SFT、LoRA、DPO等完整流程 https://github.com/hans0809/MiniMind-in-Depth * 🚀 让我们来从头训练一个属于自己的Mini-RWKV-7吧~ 小小的鹅也能飞得很高喔~ https://github.com/Alic-Li/Mini_RWKV_7 * 【DataWhale】📚 大语言模型原理与实践教程 https://github.com/datawhalechina/happy-llm * 【DataWhale】📚 智能体原理与实践教程 https://github.com/datawhalechina/hello-agents * https://github.com/jingyaogong/minimind/discussions/618 * https://github.com/jingyaogong/minimind/discussions/704 * MiniMind文档 https://minimind.readthedocs.io https://jingyaogong.github.io/minimind * 基于国产显卡华为昇腾910B复现，并且增加function calling和mcp训练数据集。 https://github.com/fzkun/minimind-ascend * 从0开始训练一个大模型Minimind项目的超详细解析，包括但不限于用到的架构，算法，以及大模型面试经验 https://github.com/Tongyun1/from-minimind-to-more 一些推荐内容，欢迎追加～

(1 comment) (18 reactions) (0 assignees)Python (49,852 stars) (6,338 forks)batch import

documentationenhancementgood first issue

Description

MiniMind的源码解读，包含tokenizer、RoPE、MoE、KV Cache、pretraining、SFT、LoRA、DPO等完整流程 https://github.com/hans0809/MiniMind-in-Depth
🚀 让我们来从头训练一个属于自己的Mini-RWKV-7吧~ 小小的鹅也能飞得很高喔~ https://github.com/Alic-Li/Mini_RWKV_7
【DataWhale】📚 大语言模型原理与实践教程 https://github.com/datawhalechina/happy-llm
【DataWhale】📚 智能体原理与实践教程 https://github.com/datawhalechina/hello-agents
https://github.com/jingyaogong/minimind/discussions/618
https://github.com/jingyaogong/minimind/discussions/704
MiniMind文档 https://minimind.readthedocs.io https://jingyaogong.github.io/minimind
基于国产显卡华为昇腾910B复现，并且增加function calling和mcp训练数据集。 https://github.com/fzkun/minimind-ascend
从0开始训练一个大模型Minimind项目的超详细解析，包括但不限于用到的架构，算法，以及大模型面试经验 https://github.com/Tongyun1/from-minimind-to-more

一些推荐内容，欢迎追加～

Contributor guide

Tech stack: pythonpytorch
Domain: documentationmachine learning
Issue type: documentation
Difficulty: 1
Estimated time: under 1 hour
Activity status: fresh
Clarity: clear
Prerequisites: None
Newbie friendliness: 95
Research direction: This issue is a curated collection of external resources (repos, discussions, documentation) related to the MiniMind project. It serves as a reference list and does not require any action from contributors. No code changes or research are needed; it simply aggregates existing content for community awareness.