I installed 1.13.0+cu117. I’m getting this warning:
NVIDIA H100 80GB HBM3 with CUDA capability sm_90 is not compatible with the current PyTorch installation.
I have to use 1.13.0. Is there a way to get it to work with H100s?
I installed 1.13.0+cu117. I’m getting this warning:
NVIDIA H100 80GB HBM3 with CUDA capability sm_90 is not compatible with the current PyTorch installation.
I have to use 1.13.0. Is there a way to get it to work with H100s?
No, since PyTorch=1.13.x supported CUDA<=11.7 while Hopper support landed in CUDA 11.8+.
PyTorch==2.0 was the next release which supported Hopper.
What about this: Hopper Architecture Compatibility
Also the compatibility matrix here seems to say that Hopper is compatible with 11.7
CUDA applications built using CUDA Toolkit versions 2.1 through 11.7 are compatible with Hopper GPUs as long as they are built to include PTX versions of their kernels. This can be tested by forcing the PTX to JIT-compile at application load time with following the steps:
You can check if PTX was shipped in these binaries and could try to JIT compile it, but I haven’t tried it using these wheels.