[BUG]: Unable to train on H20 machine

[kaixinbear]

Is there an existing issue for this bug?

• I have searched the existing issues

🐛 Describe the bug

I want to use nvidia H20 machine to run the experiment based on ColossalAI. However, I find it hard to get into the forward function of the neural module. The log reports as below:

Could anyone do me a favor ?

Environment

torch2.3 cu121
and I build the colossalai using the command:
BUILD_EXT=1 pip install colossalai==0.3.7 --no-cache-dir

[wangbluo]

Could you please offer more information? like other error message, your command and script.