Skip to content

DLESyM checkpoint compat fixes - #1949

Merged
pzharrington merged 1 commit into
NVIDIA:mainfrom
pzharrington:dlesym-checkpoint-compat
Aug 25, 2026
Merged

pzharrington merged 1 commit into
NVIDIA:mainfrom
pzharrington:dlesym-checkpoint-compat

Conversation

@pzharrington

@pzharrington pzharrington commented Aug 25, 2026 •

Copy link
Copy Markdown
Collaborator

PhysicsNeMo Pull Request

Description

#1827 actually broke the previous dlesym checkpoints (dlesym-v1-era5-v1.0 and dlesym-v0-isccp-era5). These fixes add some backward compat shims to checkpoint load so older checkpoints of previous dlesym variants will still run.

Checklist

Dependencies

Review Process

All PRs are reviewed by the PhysicsNeMo team before merging.

Depending on which files are changed, GitHub may automatically assign a maintainer for review.

We are also testing AI-based code review tools (e.g., Greptile), which may add automated comments with a confidence score.
This score reflects the AI’s assessment of merge readiness and is not a qualitative judgment of your work, nor is
it an indication that the PR will be accepted / rejected.

AI-generated feedback should be reviewed critically for usefulness.
You are not required to respond to every AI comment, but they are intended to help both authors and reviewers.
Please react to Greptile comments with 👍 or 👎 to provide feedback on their accuracy.

@pzharrington pzharrington self-assigned this Aug 25, 2026
@copy-pr-bot

copy-pr-bot Bot commented Aug 25, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@github-actions

Copy link
Copy Markdown
Contributor

CODEOWNERS review map

Current for commit 61e1fe1caa96. An approval covers every file listed for that owner; one owner is sufficient for shared files.

⏳ @pzharrington — 4 file(s)
  • physicsnemo/models/dlwp_healpix/HEALPixRecUNet.py
  • physicsnemo/models/dlwp_healpix/HEALPixUNet.py
  • physicsnemo/models/dlwp_healpix/layers/init.py
  • physicsnemo/models/dlwp_healpix/layers/healpix_blocks.py

Comment /codeowners-info to refresh.

@greptile-apps

greptile-apps Bot commented Aug 25, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

The PR restores compatibility with older DLESyM checkpoints by versioning both HEALPix U-Net variants, rewriting legacy constructor configurations, preserving the historical residual-skip rule, and backfilling a duplicated activation buffer during state loading.

  • Advances the HEALPix model checkpoint version to 0.3.0 and maps versions 0.1.0 and 0.2.0 forward.
  • Detects pre-v1.1 constructor arguments and restores legacy skip and diagnostic behavior.
  • Adds a block-level compatibility flag and tolerant loading for legacy activation-cap state keys.

Important Files Changed

Filename Overview
physicsnemo/models/dlwp_healpix/HEALPixRecUNet.py Advances checkpoint versioning and applies the shared legacy argument mapper for recurrent HEALPix checkpoints.
physicsnemo/models/dlwp_healpix/HEALPixUNet.py Advances checkpoint versioning and applies the shared legacy argument mapper for non-recurrent HEALPix checkpoints.
physicsnemo/models/dlwp_healpix/layers/init.py Adds structurally gated legacy argument rewriting; behavior is consistent with historical construction, but two new helper docstrings violate repository formatting requirements.
physicsnemo/models/dlwp_healpix/layers/healpix_blocks.py Adds the selectable historical skip rule and activation-cap state backfill; the compatibility behavior is coherent, with one new docstring using the wrong string prefix.

Reviews (1): Last reviewed commit: "dlesym checkpoint compat" | Re-trigger Greptile

Comment thread physicsnemo/models/dlwp_healpix/layers/__init__.py
Comment thread physicsnemo/models/dlwp_healpix/layers/__init__.py
@pzharrington

Copy link
Copy Markdown
Collaborator Author

/ok to test 61e1fe1

@pzharrington pzharrington changed the title DLESyM DLESyM checkpoint compat fixes Aug 25, 2026
@pzharrington
pzharrington added this pull request to the merge queue Aug 25, 2026
Merged via the queue into NVIDIA:main with commit ff4cad2 Aug 25, 2026
16 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants