{"full_name":"jingyaogong/minimind","owner":"jingyaogong","name":"minimind","url":"https://github.com/jingyaogong/minimind","description":"🧠 Train a 64M-parameter LLM from scratch in just 2h!","homepage":"https://jingyaogong.github.io/minimind","topics":[],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/62287848?u=d4a009e1468818967e9bed23e8c32b92edd2b502&v=4","created_at":"2024-07-27T04:34:12Z","pushed_at":"2026-10-27T06:57:09Z","metrics":{"stars":65087,"forks":8530,"watchers":289,"issues":660,"prs":154,"releases":2,"issues_open":null,"issues_closed":null,"prs_open":null,"prs_closed":null,"prs_merged":null,"discussions":52},"growth":{"day":{"stars":45,"forks":11,"base_day":"2026-11-09"}},"history":{"dates":["2026-11-09","2026-11-10"],"stars":[65042,65087],"forks":[8519,8530],"watchers":[289,289],"issues_open":[null,null],"prs_open":[null,null]},"boards":{},"same_owner":[{"full_name":"jingyaogong/minimind-v","name":"minimind-v","boards":{}}],"issues":[{"number":26,"title":"不是Issue，一点个人训练minimind的记录","comments":29,"created_at":"2024-09-14T09:53:01Z"},{"number":620,"title":"如何让模型学会100内的加减法？","comments":20,"created_at":"2026-01-07T07:44:46Z"},{"number":73,"title":"学习代码的时候写了一个教程，希望可以帮到其他同学","comments":9,"created_at":"2024-10-23T07:13:50Z"},{"number":97,"title":"做这个项目的一些实验结果和见解","comments":7,"created_at":"2024-12-14T13:07:46Z"},{"number":786,"title":"[not issue] 训练实验数据分享","comments":6,"created_at":"2026-05-30T13:43:15Z"},{"number":40,"title":"请教如何调整model的参数配置","comments":6,"created_at":"2024-09-22T01:05:46Z"},{"number":526,"title":"使用腾讯云Cloud Studio免费算力一键训练minimind","comments":5,"created_at":"2025-11-07T03:34:41Z"},{"number":436,"title":"感谢大佬的开源教程！这是我在学习 MiniMind[最新版本] 过程中记录的12篇笔记，包含了原理解释和代码shape注释。","comments":5,"created_at":"2025-06-17T02:01:39Z"}],"discussions":[{"number":900,"title":"[已撤回] 36M 循环深度模型分享（作者撤回）","comments":0,"category":"Show and tell","created_at":"2026-11-04T15:35:52Z"},{"number":892,"title":"是否有交流群，非常希望加入","comments":0,"category":"General","created_at":"2026-10-20T02:20:59Z"},{"number":885,"title":"be faster to get the training result with triton lib","comments":0,"category":"General","created_at":"2026-10-09T14:12:13Z"},{"number":849,"title":"[讨论] apply_lora 用 in_features == out_features 选层，实际只挂到 q_proj / o_proj","comments":0,"category":"Ideas","created_at":"2026-09-07T00:33:26Z"},{"number":848,"title":"[讨论] torch rollout 下 GRPO/CISPO 的 importance ratio 恒等于 1，clip 分支从不生效","comments":1,"category":"Ideas","created_at":"2026-09-07T00:33:26Z"},{"number":839,"title":"Measurement: error bars on the objective benchmark table (0 of 6 dense-vs-MoE differences are significant)","comments":0,"category":"General","created_at":"2026-09-03T09:14:59Z"},{"number":838,"title":"Measurement: expert utilisation in minimind-3-moe (routing is ~3.8x more imbalanced than random init)","comments":0,"category":"General","created_at":"2026-09-03T09:01:20Z"},{"number":806,"title":"GRPO 训练中生成num_generations 个回答，是如何保证回答的多样性和不重复的呢？","comments":1,"category":"Q&A","created_at":"2026-07-08T02:31:18Z"}],"latest_release":{"name":"MiniMind Docs","tag":"v2","published_at":"2025-10-21T14:37:05Z","body":"https://minimind.readthedocs.io"},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-10T09:35:25Z","history_start":"2026-09-06"}