Skip to content

setup.py: gate GroupNorm SM targets by CUDA version; avoid sm_100 #1928

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
juyterman1000 wants to merge 1 commit into
base: master
Choose a base branch
from
Open

setup.py: gate GroupNorm SM targets by CUDA version; avoid sm_100 #1928

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 11 additions & 5 deletions setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -532,11 +532,17 @@ def check_cudnn_version_and_warn(global_option: str, required_cudnn_version: int

# CUDA group norm supports from SM70
arch_flags = []
# FIXME: this needs to be done more cleanly
for arch in [70, 75, 80, 86, 90, 100, 120]:
arch_flag = f"-gencode=arch=compute_{arch},code=sm_{arch}"
arch_flags.append(arch_flag)
arch_flags.append(arch_flag)
# Gate architectures by available CUDA version to avoid unsupported targets on older toolkits
allowed_arches = [70, 75, 80]
if bare_metal_version >= Version("11.1"):
allowed_arches.append(86)
if bare_metal_version >= Version("11.8"):
allowed_arches.append(90)
if bare_metal_version >= Version("12.8"):
allowed_arches.extend([100, 120])

for arch in allowed_arches:
arch_flags.append(f"-gencode=arch=compute_{arch},code=sm_{arch}")

ext_modules.append(
CUDAExtension(
Expand Down