Fix DeepCompile profile metadata backfill - #8094
Merged
tohtana merged 7 commits intoJun 27, 2026
Merged
Conversation
Copy link
Copy Markdown
Collaborator
Sorry, something went wrong.
DeepCompile ZeRO-3 scheduler can crash with KeyError or AssertionError on missing device_time when graph profiling fails partway. Backfill default timing and memory metadata for incomplete profile nodes so the scheduler and selective_gather pass remain robust. Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
Copy link
Copy Markdown
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3f2013f11a
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Sorry, something went wrong.
Comment thread
deepspeed/compile/profilers/graph_profile.py
Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
Copy link
Copy Markdown
There was a problem hiding this comment.
Pull request overview
This PR hardens DeepCompile graph profiling so downstream scheduling/passes (notably ZeRO-3 scheduling and selective gather) won’t crash when profiling aborts partway through and leaves some FX nodes without timing/memory metadata.
Changes:
- Add a post-profiling backfill step to ensure every FX node has default profile metadata keys (device_time, wall_time, tensor_size, alloc_mem, max_mem) even after profiling errors/early exits.
- Make selective-gather’s size/time ranking more robust when profiled tensor sizes are missing/zero by avoiding overwriting param-derived sizes and avoiding divide-by-zero.
- Add a unit test that simulates partial profiling metadata and validates both scheduling and selective-gather behavior remain safe.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| deepspeed/compile/profilers/graph_profile.py | Introduces metadata defaults and backfills missing node metadata in finally to guarantee completeness after profiling cleanup/errors. |
| deepspeed/compile/passes/selective_gather.py | Prevents invalid/zero profiled sizes from clobbering param-derived sizes; avoids divide-by-zero in time-per-byte ranking. |
| tests/unit/compile/test_list_schedule.py | Adds regression test covering partial profile metadata backfill and ensuring scheduler + selective gather don’t crash. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Sorry, something went wrong.
Comment thread
deepspeed/compile/passes/selective_gather.py
Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
Comment thread
deepspeed/compile/passes/selective_gather.py
Comment thread
deepspeed/compile/profilers/graph_profile.py
Comment thread
deepspeed/compile/profilers/graph_profile.py
Outdated
sfc-gh-truwase
approved these changes
Jun 27, 2026
Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
tohtana
enabled auto-merge
June 27, 2026 17:51
This was referenced Jun 28, 2026
nathon-lee
pushed a commit
to nathon-lee/DeepSpeed_woo
that referenced
this pull request
Jul 1, 2026
DeepCompile ZeRO-3 scheduling can crash with KeyError or AssertionError on missing device_time when graph profiling fails partway and leaves some nodes without timing metadata. This PR backfills default timing and memory metadata after profiling cleanup so the profiler guarantees downstream passes see complete metadata on every node. --------- Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com> Signed-off-by: nathon-lee <leejianwoo@gmail.com>
nathon-lee
pushed a commit
to nathon-lee/DeepSpeed_woo
that referenced
this pull request
Jul 1, 2026
DeepCompile ZeRO-3 scheduling can crash with KeyError or AssertionError on missing device_time when graph profiling fails partway and leaves some nodes without timing metadata. This PR backfills default timing and memory metadata after profiling cleanup so the profiler guarantees downstream passes see complete metadata on every node. --------- Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Reviewers
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
DeepCompile ZeRO-3 scheduling can crash with KeyError or AssertionError on missing device_time when graph profiling fails partway and leaves some nodes without timing metadata.
This PR backfills default timing and memory metadata after profiling cleanup so the profiler guarantees downstream passes see complete metadata on every node.