Cannot use AT_CUDA_DRIVER_CHECK from user code #98587
Labels
module: build
Build system issues
module: cpp-extensions
Related to torch.utils.cpp_extension
module: cuda
Related to torch.cuda, and CUDA support in general
module: internals
Related to internal abstractions in c10 and ATen
triaged
This issue has been looked at a team member, and triaged and prioritized into an appropriate module
馃悰 Describe the bug
In my C++ extension that links against PyTorch I cannot use the AT_CUDA_DRIVER_CHECK macro to check the return code of a call to the CUDA driver API, because that macro makes use of the at::cuda::NVRTC struct which is defined in the ATen/cuda/nvrtc_stub/ATenNVRTC.h header, which doesn't get installed/shipped by PyTorch!
Versions
I'm using PyTorch 2.0.0 installed from the official conda channels (build py3.9_cuda11.7_cudnn8.5.0_0).
cc @malfet @seemethere @zou3519 @ezyang @bhosmer @smessmer @ljk53 @bdhirsh @ngimel
The text was updated successfully, but these errors were encountered: