Skip to content

1.1.0: ship Sen2 SR - #132

Merged
simon-donike merged 10 commits into
mainfrom
dev
Nov 5, 2026
Merged

simon-donike merged 10 commits into
mainfrom
dev

Conversation

@simon-donike

@simon-donike simon-donike commented Nov 5, 2026 •

Copy link
Copy Markdown
Member

Adds the Sen2 SR pretrained preset, SwinIR, back-projection, and SRGAN loss/resume updates. Release 1.1.0.

Recipes, splits, checkpoints and model cards live on Hugging Face. Existing downloads keep working.

Fixes a leaking fake trainer in the tests and preserves active EMA weights when saving checkpoints.

Local validation: 528 tests passed; 96.81% project coverage, 99.26% changed-line coverage against main. Python 3.12–3.14 and both Codecov checks pass. CI configuration unchanged.

Promote the preserved real-data checkpoint to a strict, immutable Hugging Face preset with validated EMA-only export. Port the matching shuffled-band local-Gram, global RaGAN, normalized R1, multiband validation and resume support while retaining legacy inference behavior.

Validation: 341 tests passed and 1 skipped; distributed RaGAN check passed with loopback access. Real full-checkpoint restoration and one training update passed. Original experiment, full checkpoint, exported weights and cold-cache installed-wheel inference matched exactly; content losses and SR gradients also matched exactly.
Add versioned model cards, modular checkpoint/config paths, MIT licenses, tensor/GeoTIFF/resume examples, catalogs and checksums. Publish existing artifacts as server-side copies while retaining every historical artifact URL and byte sequence.

Pin all presets and sample downloads to Hub commit b31d2728952cfaa022bd1df92d82a54c23cf4f8f. Validation: 30 loader, downloader, inference and deployment tests passed; all three canonical cold-cache presets exactly match original outputs; 41 published file hashes verified; dataset extraction produced 200 files; actual 4x/8x GeoTIFF examples and resume preparation passed; wheel builds.
Finalize package and citation metadata, document the Sen2 SR release and upgrade behavior, and use stable installation commands. Clarify inference model placement and add release notes to the docs navigation.
@codecov

codecov Bot commented Nov 5, 2026 •

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 99.26335% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.84%. Comparing base (b84ca10) to head (ccc0811).
⚠️ Report is 4 commits behind head on main.

Files with missing lines Patch % Lines
opensr_srgan/model/adversarial.py 85.29% 5 Missing ⚠️
opensr_srgan/model/SRGAN.py 98.99% 2 Missing ⚠️
opensr_srgan/model/training_step_PL.py 96.55% 1 Missing ⚠️
Additional details and impacted files

Impacted file tree graph

@@            Coverage Diff             @@
##             main     #132      +/-   ##
==========================================
+ Coverage   94.58%   96.84%   +2.25%     
==========================================
  Files          32       38       +6     
  Lines        1940     2949    +1009     
==========================================
+ Hits         1835     2856    +1021     
+ Misses        105       93      -12     
Files with missing lines Coverage Δ
opensr_srgan/__init__.py 100.00% <ø> (ø)
opensr_srgan/_factory.py 100.00% <100.00%> (ø)
opensr_srgan/model/generators/__init__.py 100.00% <100.00%> (ø)
opensr_srgan/model/generators/factory.py 96.51% <100.00%> (+0.17%) ⬆️
opensr_srgan/model/generators/swinir.py 100.00% <100.00%> (ø)
opensr_srgan/model/loss/local_gram.py 100.00% <100.00%> (ø)
opensr_srgan/model/loss/loss.py 90.78% <100.00%> (+6.80%) ⬆️
opensr_srgan/model/loss/vgg.py 100.00% <100.00%> (ø)
opensr_srgan/pretrained.py 100.00% <100.00%> (ø)
opensr_srgan/train.py 98.09% <100.00%> (+4.68%) ⬆️
... and 8 more

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@simon-donike
simon-donike merged commit 1f5b06d into main Nov 5, 2026
7 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.

1 participant