InternLM/InternEvo

View on GitHub

InternEvo is an open-sourced lightweight training framework aims to support model pre-training without the need for extensive dependencies.

Stars: 421
Forks: 67
Open beginner issues: 0
Indexed issues: 49

Dominant language: Python
License: Apache-2.0
Last GitHub push: Aug 21, 2025
Latest indexed: Sep 17, 2026
Contributing guide: No contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed
PR merge metrics: No merged PRs in 30d

Open All

49 open issues indexed

[Feature] 放松依赖版本限制 Open

enhancement

InternLM/InternEvo#14 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] Partially frozen model support Open

enhancement

InternLM/InternEvo#11 · 2 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] 请在pipeline parallel上支持不同长度的input, label的计算。 Open

enhancement

InternLM/InternEvo#9 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] rotary代码规范化 Open

enhancement

InternLM/InternEvo#33 · 0 comments · 0 reactions · 3 assignees · Indexed Sep 17, 2026
[Typo] `schedulder` -> `scheduler` Open

question

InternLM/InternEvo#49 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] add internlm2-1.8b finetuning config Open

enhancement

InternLM/InternEvo#71 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
升级CUDA版本以支持Windows版本的flash-attention Open

enhancement

InternLM/InternEvo#78 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
在安装docker环境时，总是爆出这个错误，无法解决 Open

bug

InternLM/InternEvo#85 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] support sequence parallel in head layer and embedding layer Open

enhancement

InternLM/InternEvo#95 · 1 comment · 0 reactions · 2 assignees · Indexed Sep 17, 2026
[Feature] CI should have a true no flashattention env Open

bug enhancement

InternLM/InternEvo#96 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] a very simple hugging-face dataloader Open

enhancement

InternLM/InternEvo#101 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] supporting hugging-face modeling python file Open

enhancement

InternLM/InternEvo#109 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] update readme with new version of dependency. Open

enhancement

InternLM/InternEvo#121 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] CPU synchronization Problem Open

enhancement

InternLM/InternEvo#122 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] do not use torch.cuda.current_device() as device, since it only retures an int Open

bug

InternLM/InternEvo#130 · 3 comments · 0 reactions · 2 assignees · Indexed Sep 17, 2026
[Bug] the usage "x.is_cuda" is not recommended, since we have both GPU and NPU Open

bug

InternLM/InternEvo#134 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] Recommend to replace internlm1.1 with internlm2 in web-demo Open

enhancement

InternLM/InternEvo#144 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] Support customized model size for training Open

enhancement

InternLM/InternEvo#152 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] 使用moe的config微调报错 Open

bug

InternLM/InternEvo#153 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] Should we remove other dependency of flashattention? Open

enhancement

InternLM/InternEvo#164 · 0 comments · 0 reactions · 4 assignees · Indexed Sep 17, 2026
[QA] 关于使用张量并行或流水线并行的模型切分与合并问题 Open

question

InternLM/InternEvo#176 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] 昇腾910微调internLM报错 Open

bug

InternLM/InternEvo#212 · 3 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
Add tool for data cleaning Open

question

InternLM/InternEvo#241 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] AssertionError: Only flash cross entropy support parallel_output Open

bug

InternLM/InternEvo#245 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] 用Internevo已经训练出来了一个7B模型，如何用这个internevo权重跑MoE？ Open

question

InternLM/InternEvo#251 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] Internevo这个框架里面MoE支持expert parallel嘛？ Open

question

InternLM/InternEvo#253 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] 好像没有把internevo的MoE权重转换成huggingface版本的脚本？ Open

bug

InternLM/InternEvo#262 · 8 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] 用MoE训练的时候tflop超级低 Open

bug

InternLM/InternEvo#264 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
[Bug] 使用internevo训练，转换成hf模型用opencompass测试时候有一定概率会nan Open

bug

InternLM/InternEvo#266 · 2 comments · 0 reactions · 2 assignees · Indexed Sep 17, 2026
[QA] Internevo是否支持tied_embedding? Open

question

InternLM/InternEvo#267 · 3 comments · 0 reactions · 2 assignees · Indexed Sep 17, 2026
[QA] InternEvo能否load预训练llama2的参数 Open

question

InternLM/InternEvo#268 · 4 comments · 0 reactions · 2 assignees · Indexed Sep 17, 2026
[Bug] TFLOPS计算不准 Open

bug

InternLM/InternEvo#269 · 2 comments · 0 reactions · 2 assignees · Indexed Sep 17, 2026
[Bug] 训练bf16 infer fp16出现NaN Open

bug

InternLM/InternEvo#278 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] 仅支持了GShard模式的MoE模型转huggingface Open

bug

InternLM/InternEvo#279 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] RuntimeError: [3] is setting up NCCL communicator and retrieving ncclUniqueId from [0] via c10d key-value store by key '0', but store->get('0') got error: Socket Timeout Open

bug

InternLM/InternEvo#285 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] how to finetuning lora Open

enhancement

InternLM/InternEvo#286 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] check import system var at the start of training Open

question

InternLM/InternEvo#300 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] 不使用memory pool Open

enhancement

InternLM/InternEvo#324 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] MoE模型里稠密层和专家层zero和并行的解耦 Open

enhancement

InternLM/InternEvo#325 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] loong train 支持packed_sample_into_one=false吗 Open

question

InternLM/InternEvo#346 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] 如何进行单卡微调的，需要调整那些设置 Open

question

InternLM/InternEvo#354 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] Does internEvo support loongtrain selective checkpoint++? Open

question

InternLM/InternEvo#372 · 1 comment · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[QA] 训练到一半Nan grad norm occurs, please check it Open

question

InternLM/InternEvo#375 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] loss error: flat_fp32_avg_grads will not be scaled when hybridzerooptimizer use cpu_offload Open

bug

InternLM/InternEvo#398 · 2 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] Add Lingen - MATE architecture (SSM Mamba2 based) as a model option Open

enhancement

InternLM/InternEvo#402 · 2 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] TypeError: Qwen2.__init__() got an unexpected keyword argument 'enable_qkv_fusion' Open

bug

InternLM/InternEvo#408 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] CPU Mem utilization grows with training, when Dataloader num_workers>0 Open

bug

InternLM/InternEvo#415 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Bug] Some spelling errors in repo Open

bug

InternLM/InternEvo#417 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026
[Feature] 该怎么在VLM训练里也用这个框架？有示例吗 Open

enhancement

InternLM/InternEvo#426 · 0 comments · 0 reactions · 1 assignee · Indexed Sep 17, 2026

InternLM/InternEvo

Get new issues in your inbox