Spaces:

---

BestWishYsh

/

Helios-14B-RealTime-AOTI

Like 83

Runtime error

Hi! Thanks for working on the restore.
The Space is currently crashing at startup with a Runtime Error. It looks like the deployment fails during the compilation of Torch Inductor / AOTInductor on ZeroGPU.
Here is the key part of the traceback:
fatal error: cuda.h: No such file or directory
gradio.exceptions.Error: "CppCompileError"
It seems the build environment is missing the CUDA Toolkit headers required for the C++ compilation step at runtime, or ZeroGPU is blocking the dynamic compilation.
Could you please check the Docker/environment configuration or look into disabling/pre-compiling the Torch Inductor wrapper?
Thank you!