Skip to content

Trim non-pipeline input grads before caching in bwd_cache #431

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
aditvenk wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Trim non-pipeline input grads before caching in bwd_cache #431

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
3 changes: 3 additions & 0 deletions autoparallel/graph_passes/graph_pp_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -672,6 +672,9 @@ def _post_backward_common(
stage_index_to_stage: Dictionary mapping stage indices to GraphPipelineStage objects.
is_prev_stage_on_this_rank: True if the previous stage exists on this rank.
"""
assert bw_stage._stage_meta.inputs is not None
num_fwd_args = len(bw_stage._stage_meta.inputs)
input_grads = input_grads[:num_fwd_args]
bw_stage.bwd_cache[bw_mb_index] = (
tuple(input_grads) if not isinstance(input_grads, tuple) else input_grads
)
Expand Down
Loading