github
/
minimind
mirror of https://github.com/jingyaogong/minimind.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • efd8073450 adapt to Ascend NPU yanghw116 2026-05-18 18:36:14 +0800
  • 182bff1bd8 fix: guard against empty/filtered LLM responses in chat_api.py Jonathan Hill 2026-05-18 03:44:31 -0500
  • e90f92d701 [update] requirements jingyaogong 2026-05-16 18:08:37 +0800
  • 9da8e1ab18 [update] requirements jingyaogong 2026-05-16 18:08:37 +0800
  • 579cc79cd8 fix: extract tools and tool_calls before apply_chat_template in PPL eval baihaochen 2026-05-16 11:33:55 +0800
  • 15645bbf56 [feat] add GQA vs MLA evaluation script (PPL + Q&A) baihaochen 2026-05-16 11:27:49 +0800
  • 7227c6a21c chore(args): 支持Mac MPS设备作为运行和训练设备 广峰 2026-05-16 01:26:42 +0800
  • 0074dcfd50
         Merge branch 'jingyaogong:master' into master 100apps 2026-05-15 22:56:52 +0800
  • 09e491f940 Merge branch 'master' of github.com:daiweidaichenzi/minimind baihaochen 2026-05-14 14:29:06 +0800
  • 8baa9c5f71 fix: add missing --use_mla to train_full_sft.py baihaochen 2026-05-14 14:28:11 +0800
  • de683eafae merge local changes rainpeach-m 2026-05-12 11:18:45 +0000
  • 49a05a1307 增加mla注意力模型 baihaochen 2026-05-12 18:57:53 +0800
  • bc3d137248 [feat] add MLA with decoupled RoPE (DeepSeek-V2 style) baihaochen 2026-05-12 18:38:47 +0800
  • 743420daac [fix] 更新.gitignore文件，添加忽略目录；修改web_demo.py，优化样式和功能 Toward 2026-05-10 23:09:26 +0800
  • 9c23a92922 add Chinese annotations pretrain baihaochen 2026-05-08 20:23:09 +0800
  • 4b0993fcc2 [fix] repetition_penalty jingyaogong 2026-05-07 19:08:52 +0800
  • dddedc6881 [fix] repetition_penalty jingyaogong 2026-05-07 19:08:52 +0800
  • a22f1f1486 [feat] reduce RL memory jingyaogong 2026-05-06 15:07:28 +0800
  • 802c15b2b4 [feat] reduce RL memory jingyaogong 2026-05-06 15:07:28 +0800
  • ed7cbcee19 Merge pull request #759 from TKiteRunner/fix/grpo-oom-reorder-rewards jingyaogong 2026-05-06 15:00:17 +0800
  • e73a407f7a
         Merge pull request #759 from TKiteRunner/fix/grpo-oom-reorder-rewards jingyaogong 2026-05-06 15:00:17 +0800
  • 1c583deb28 fix: resolve CUDA OOM in train_grpo.py on GPUs with <=8GB VRAM TKiteRunner 2026-05-06 14:14:15 +0800
  • 10776417aa fix: resolve CUDA OOM in train_grpo.py on GPUs with <=8GB VRAM TKiteRunner 2026-05-06 14:14:15 +0800
  • 43eee6add5 [update] readme jingyaogong 2026-05-06 13:41:46 +0800
  • 5020dc9dd4 [update] readme jingyaogong 2026-05-06 13:41:46 +0800
  • 68c00c22b0
         Fix path separator in download command 100apps 2026-05-06 12:42:34 +0800
  • ebf36783b8 fix(eval_toolcall): 修复模型路径解析与加载问题 100apps 2026-05-05 21:34:10 +0800
  • 7cf7881064 docs(readme): 优化已训练模型测试说明 100apps 2026-05-05 18:39:37 +0800
  • 509d2522ba docs(readme): 优化已训练模型测试说明 100apps 2026-05-05 18:37:38 +0800
  • 5fc34ff6c5 docs(readme): 补充模型权重下载命令示例 100apps 2026-05-05 18:36:03 +0800
  • b9de7c17fb docs(readme): 优化数据集下载说明部分格式 100apps 2026-05-05 17:47:14 +0800
  • 8607169546 fix: V-001 security vulnerability orbisai0security 2026-05-05 07:51:07 +0000
  • 97903ffe56 chore(deps): 使用uv pyproject.toml替代requirements.txt管理依赖并增加vscode debug 配置，方便debug。 100apps 2026-05-05 15:26:33 +0800
  • 58959684c7 [fix] inference bug jingyaogong 2026-05-03 20:48:48 +0800
  • bdee223036 [fix] inference bug jingyaogong 2026-05-03 20:48:48 +0800
  • e6435db723 提交到github Schenyu 2026-05-02 16:48:16 +0800
  • 91bb73767b 加一个测试文件 Schenyu 2026-05-02 15:52:29 +0800
  • 941c39c3ec 测试下git上传 Schenyu 2026-05-02 15:28:59 +0800
  • e58fb8269f first commit Schenyu 2026-05-02 15:16:05 +0800
  • d1148341bb [update] readme jingyaogong 2026-05-01 12:06:47 +0800
  • 06d882e4ef [update] readme jingyaogong 2026-05-01 12:06:47 +0800
  • 7b4f029ee3 [update] readme jingyaogong 2026-04-28 17:22:25 +0800
  • da865af63d [update] readme jingyaogong 2026-04-28 17:22:25 +0800
  • 4b85ba8b28 [fix] bugs jingyaogong 2026-04-27 19:16:08 +0800
  • 773e451b11 [fix] bugs jingyaogong 2026-04-27 19:16:08 +0800
  • 3b7a4d04a2 [fix] rollout bugs jingyaogong 2026-04-27 17:54:09 +0800
  • 6361510016 [fix] rollout bugs jingyaogong 2026-04-27 17:54:09 +0800
  • b45c36b8b5 [update] readme jingyaogong 2026-04-27 10:59:41 +0800
  • d4c6bc5c7e [update] readme jingyaogong 2026-04-27 10:59:41 +0800
  • e4e9153a1f [update] readme jingyaogong 2026-04-24 20:16:06 +0800
  • 24896cd2c4 [update] readme jingyaogong 2026-04-24 20:16:06 +0800
  • 29c87cfc8d [update] readme jingyaogong 2026-04-24 20:14:10 +0800
  • e2fc397176 [update] readme jingyaogong 2026-04-24 20:14:10 +0800
  • d619d86c4c
         fix: replace dangerous expression evaluation with AST-validated safe math parser (CWE-95) Sebastion 2026-04-23 19:36:39 +0100
  • c0eb98393f [update] readme jingyaogong 2026-04-21 14:34:46 +0800
  • 693fb1ccf1 [update] readme jingyaogong 2026-04-21 14:34:46 +0800
  • 568a863520 [fix] bugs jingyaogong 2026-04-21 13:03:34 +0800
  • 5416a44471 [fix] bugs jingyaogong 2026-04-21 13:03:34 +0800
  • 761710654a [fix] transformers-5.x jingyaogong 2026-04-19 23:48:54 +0800
  • 1718e9a44d [fix] transformers-5.x jingyaogong 2026-04-19 23:48:54 +0800
  • 145a7ac6f6 [update] tie embedding jingyaogong 2026-04-19 21:57:28 +0800
  • 5704766352 [update] tie embedding jingyaogong 2026-04-19 21:57:28 +0800
  • 83b4c697a3 [update] readme jingyaogong 2026-04-19 14:51:47 +0800
  • 1ea113ea2c [update] readme jingyaogong 2026-04-19 14:51:47 +0800
  • 48ed5ec8bc fix: OpenAI API SSE compatibility and stream reliability voidborne-d 2026-04-17 00:56:36 +0000
  • ef40a1f271 一些小的改动 翟锦洋 2026-04-12 18:06:52 +0800
  • b9a229e1ae [update] readme jingyaogong 2026-04-10 10:55:24 +0800
  • 487f78754d [update] readme jingyaogong 2026-04-10 10:55:24 +0800
  • e612cc26e8 [fix] lora compile jingyaogong 2026-04-10 00:01:21 +0800
  • e796b8028a [fix] lora compile jingyaogong 2026-04-10 00:01:21 +0800
  • 903eb1dd37 [update] readme jingyaogong 2026-04-09 19:00:41 +0800
  • b2488e6440 [update] readme jingyaogong 2026-04-09 19:00:41 +0800
  • 6b9588d619 [update] readme jingyaogong 2026-04-09 18:54:17 +0800
  • 939dc8ff42 [update] readme jingyaogong 2026-04-09 18:54:17 +0800
  • d942f24bf1 [update] bench jingyaogong 2026-04-09 17:12:48 +0800
  • cadacabecb [update] bench jingyaogong 2026-04-09 17:12:48 +0800
  • 862dd151f2 [fix] lora moe jingyaogong 2026-04-09 16:36:48 +0800
  • 5351424bf0 [fix] lora moe jingyaogong 2026-04-09 16:36:48 +0800
  • b703844a60 [update] add a comment jingyaogong 2026-04-09 15:09:58 +0800
  • aa3e6affa1 [update] add a comment jingyaogong 2026-04-09 15:09:58 +0800
  • d399015c04 [update] ignore local datasets and model artifacts wzz 2026-04-05 18:18:48 +0000
  • cfa8247f20 test wzz 2026-04-05 18:17:05 +0000
  • d37bfa9d75 [update] harden training and inference reliability root 2026-04-05 18:11:37 +0000
  • 299facca84 [update] image website jingyaogong 2026-04-04 14:55:03 +0800
  • 428f017e21 [update] image jingyaogong 2026-04-04 14:54:13 +0800
  • 25a7edcd6f [update] image jingyaogong 2026-04-04 14:54:13 +0800
  • ef1c130a03 [update] readme jingyaogong 2026-04-04 11:25:21 +0800
  • cacf1d4cd0 [update] readme jingyaogong 2026-04-04 11:25:21 +0800
  • 367838379a fix: repetition_penalty boosts negative-logit tokens instead of suppressing them d 🔹 2026-04-03 12:11:16 +0000
  • 7546729893 [update] open causal jingyaogong 2026-04-02 15:28:58 +0800
  • 2ab6455d9d [update] open causal jingyaogong 2026-04-02 15:28:58 +0800
  • 6512d5647a [update] readme jingyaogong 2026-04-02 15:28:29 +0800
  • 9348fde743 [update] readme jingyaogong 2026-04-02 15:28:29 +0800
  • ea57409fd2 [update] readme jingyaogong 2026-04-01 14:00:21 +0800
  • 90cd275524 [update] readme jingyaogong 2026-04-01 14:00:21 +0800
  • d52691f2e3 [update] default model jingyaogong 2026-03-31 13:40:16 +0800
  • b7e0ae21d6 [update] default model jingyaogong 2026-03-31 13:40:16 +0800
  • 5e9d15614e [update] random seed jingyaogong 2026-03-27 21:20:02 +0800
  • b1865f75c2 [update] random seed jingyaogong 2026-03-27 21:20:02 +0800
  • 40358702ab 使用einops进一步提升代码可读性 wizardforcel 2026-03-27 19:22:53 +0800