1 Pull request opened by 1 person

• feat: 优化导入和代码风格

  #162 opened Feb 11, 2025

31 Issues closed by 2 people

• 数据集链接失效

  #160 closed Feb 11, 2025

• 关于：1-pretrain.py与3-full-sft.py中的train_epoch

  #126 closed Feb 10, 2025

• Is it possible to train for a code model?

  #158 closed Feb 10, 2025

• mac book pro m1能用吗

  #127 closed Feb 10, 2025

• 考虑使用MLA吗

  #133 closed Feb 10, 2025

• 参数量大的模型会爆显存吗？有必要用megatron吗？

  #157 closed Feb 10, 2025

• 考虑加入蒙特卡洛搜索吗

  #131 closed Feb 10, 2025

• 您好，我想了解下Seq-Monkey通用文本数据集的构建

  #128 closed Feb 10, 2025

• 如何训练一个垂直领域的模型

  #122 closed Feb 10, 2025

• 考虑加一个Reasoning模型么？

  #125 closed Feb 10, 2025

• Is it possible to update the model to DeepSeek-V3 architecture?

  #120 closed Feb 10, 2025

• 关于 MoE 版本的辅助损失函数.

  #110 closed Feb 10, 2025

• 关于在MAC上执行的一些发现，非issue

  #55 closed Feb 10, 2025

• 想问一下作者大大是怎么做到模型大小这么小的，通过模型压缩或是参数共享吗，还是只是减少了layer、dim这些参数呢？

  #105 closed Feb 10, 2025

• 讨论个人GPU的训练时间

  #60 closed Feb 10, 2025

• 使用google colab方式训练测试笔记本（不是issue）

  #50 closed Feb 10, 2025

• 关于“”更高效的kvcache“”的疑问

  #156 closed Feb 10, 2025

• MoE专家的aux loss好像未被使用

  #140 closed Feb 10, 2025

• tokenizer解码再编码之后和原文不一样

  #135 closed Feb 10, 2025

• why use BOS and EOS token and not only the EOS?

  #130 closed Feb 10, 2025

• moe共享专家

  #129 closed Feb 10, 2025

• 部分网盘链接失效

  #124 closed Feb 10, 2025

• 这个只能训练中文Model吗，英文model可以训练吗，求大佬讲解

  #119 closed Feb 10, 2025

• 训练时,输入的token会补全到最大长度.那是不是意味着attention计算时,是按照满token的时间复杂度计算的

  #118 closed Feb 10, 2025

• 关于保存模型大小的问题

  #116 closed Feb 10, 2025

• 使用DPP进行多卡训练报错

  #115 closed Feb 10, 2025

• 新手请教一下大佬，我现在想让大模型完全记忆一个代码仓库的代码，我是放在预训练阶段呢呢，还是放在SFT阶段呢，或者其他阶段呢？

  #117 closed Feb 10, 2025

• Githubpage.io ✖️✖️

  #113 closed Feb 10, 2025

• 关于 tokenizer.json 词汇表中没有中文的问题

  #111 closed Feb 10, 2025

• dpo强化学习报错'generator' object has no attribute 'generate'

  #141 closed Feb 10, 2025

• 麻烦帮忙解答一下MoEFFN中训练和推理的代码底层逻辑有何不同

  #139 closed Feb 9, 2025

5 Issues opened by 5 people

• 为什么训练过程中有差不多50%是硬件是闲置的

  #164 opened Feb 12, 2025

• 数据增强(TinyStories)+自我安利

  #163 opened Feb 11, 2025

• 关于多阶段 SFT

  #161 opened Feb 11, 2025

• 如何切换训练的模型呢

  #159 opened Feb 10, 2025

• MiniMind2 is coming soon

  #142 opened Feb 6, 2025

4 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• 学习代码的时候写了一个教程，希望可以帮到其他同学

  #73 commented on Feb 9, 2025 • 0 new comments

• 长文本能力问题

  #112 commented on Feb 10, 2025 • 0 new comments

• 不是Issue，一点个人训练minimind的记录

  #26 commented on Feb 10, 2025 • 0 new comments

• 这。。。这不是我的梦中情库嘛 求大佬建个微信群~

  #9 commented on Feb 12, 2025 • 0 new comments