[Bug] GLM5模型转megatron报错

[llp1992]

Bug Description

当前提供的镜像及megatron core版本，无法支持GLM5模型hf转torch_dist，GLM5的架构还未支持。但是怎么本项目中能提供训练脚本？

Steps to Reproduce

当前提供的镜像及megatron core版本，无法支持GLM5模型hf转torch_dist，GLM5的架构还未支持。但是怎么本项目中能提供训练脚本？

Expected Behavior

正常转换GLM5

Actual Behavior

转换报错

Environment

• slime version:
• Python version:
• PyTorch version:
• CUDA/ROCm version:
• GPU type and count:
• OS:
• SGLang version (if relevant):
• Megatron-LM version (if relevant):

Logs

Additional Context

No response

Pre-submission Checklist

• I have read the CONTRIBUTING.md and understand the collaboration scope.
• I have read the documentation and my issue is not addressed there.
• I have searched for existing issues and this is not a duplicate.
• I have provided a minimal, reproducible example.

[llp1992]

@zhuzilin @Gao016

[Gao016]

@zhuzilin @Gao016

see here #1599

[Mrkkew]

最新的镜像还不支持GLM5模型吗

[aJupyter]

求一下转换脚本

[Gao016]

转换脚本可以参考这个：https://github.com/THUDM/slime/blob/main/docs/zh/examples/deepseek-r1.md

改成GLM-5对应的配置就可以

[aJupyter]

转换脚本可以参考这个：https://github.com/THUDM/slime/blob/main/docs/zh/examples/deepseek-r1.md

改成GLM-5对应的配置就可以

多谢

[aJupyter]

另外GLM5 tools/convert_torch_dist_to_hf.py 也有问题，单机8卡跑了一会直接kill了，求方案 @zhuzilin

[aJupyter]

训练中保存hf也有问题

[samaritan1998]

+1

[aJupyter]

+1

是遇到我的bug吗

[ZhijianZhou]

单机转跑不下 得两机跑

[samaritan1998]

+1

是遇到我的bug吗

是的 不过已经解决了 2机能转 需要把配置的model_type改成deepseek_v32