Skip to content

Disable dynamic graph support in Triton to prevent out-of-memory erro… #72

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

Closed
romeokienzler wants to merge 1 commit into from
Closed

Disable dynamic graph support in Triton to prevent out-of-memory erro… #72

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
6 changes: 6 additions & 0 deletions gridfm_graphkit/cli.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -161,6 +161,12 @@ def main_cli(args):
run_name=args.run_name,
)

# When using torch.compile with Triton, dynamic graph support can cause
# out-of-memory errors during autotuning on some kernels.
# Disabling dynamic graph support allows those kernels
# to be skipped gracefully instead of causing errors.
torch._inductor.config.triton.cudagraph_skip_dynamic_graphs = True

with open(args.config, "r") as f:
base_config = yaml.safe_load(f)

Expand Down
Loading