Skip to content

[simple_fsdp] Use schedule_overlap_bucketing_from_inductor_configs for overlap_passes #2346

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
IvanKobzarev wants to merge 1 commit into
base: main
Choose a base branch
from
+3 −3
Open

[simple_fsdp] Use schedule_overlap_bucketing_from_inductor_configs for overlap_passes #2346

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: 3 additions & 3 deletions torchtitan/experiments/simple_fsdp/backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ def get_compile_backend_with_passes(
# The autobucketing logic is here: https://github.com/pytorch/pytorch/pull/163960
from torch._inductor.config import aten_distributed_optimizations as dist_opts
from torch._inductor.fx_passes.overlap_scheduling import (
schedule_overlap_bucketing,
schedule_overlap_bucketing_from_inductor_configs,
)

dist_opts.collective_bucketing = True
Expand All @@ -52,7 +52,7 @@ def get_compile_backend_with_passes(
def aot_eager_autobucketing_reordering_pass(
gm: torch.fx.GraphModule, example_inputs: Any
) -> torch.fx.GraphModule:
schedule_overlap_bucketing(gm)
schedule_overlap_bucketing_from_inductor_configs(gm)
gm.recompile()
return gm

Expand All @@ -67,7 +67,7 @@ def aot_eager_autobucketing_reordering_pass(
def inductor_autobucketing_reordering_pass(
gm: torch.fx.Graph,
) -> torch.fx.GraphModule:
return schedule_overlap_bucketing(gm.owning_module)
return schedule_overlap_bucketing_from_inductor_configs(gm.owning_module)

dist_opts.insert_overlap_deps = True
torch._inductor.config.reorder_for_peak_memory = False
Expand Down
Loading