Torchgeo mlm

[Nazim-crim]

Description

• This MR introduces the from_torch function in torch/ a utility to programmatically generate a STAC Item with the Machine Learning Model (MLM) Extension from a PyTorch nn.Module.
• Add an exemple json of a parsed torchgeo unet using SENTINEL2_2CLASS_NC_FTW default weights

Related Issue

• Closes Torchgeo mlm crim-ca/mlm-extension#41
• Fixes Request inference example in pytorch/lightning #70

Type of Change

• 📚 Examples, docs, tutorials or dependencies update;
• 🔧 Bug fix (non-breaking change which fixes an issue);
• 🥂 Improvement (non-breaking change which improves an existing feature);
• 🚀 New feature (non-breaking change which adds functionality);
• 💥 Breaking change (fix or feature that would cause existing functionality to change);
• 🔐 Security fix.

Checklist

• I've read the CONTRIBUTING.md guide;
• I've updated the code style using make check;
• I've written tests for all new methods and classes that I created;
• I've written the docstring in Google format for all the methods and classes that I used.

[Nazim-crim]

Following previous comment on crim-ca/mlm-extension#41 , here's my up to date change. Let me know if other fixes are needed. Thanks

[fmigneault]

Self Note

Merge other features and release MLM 1.5.0 & stac_model 0.4.0 before merging this to avoid the potentially breaking change for users employing the package.

[fmigneault]

Self-note

Make sure the items are ordered correctly considering the intermediate release of MLM / stac_model.

[fmigneault]

@Nazim-crim
The linting check fails for the same reason as moregeo-it/stac-node-validator#78 (see the run outputs that converts 0.0 -> 0 and 3000.0 -> 3000). Since there is no planed fix for it, can you try to explicitly use int values and see if that patches the example?

The other failures for Python 3.10 should be handled properly when merging with #97.

[fmigneault]

Temp fix for STAC validation is to used stac_version: 1.0.0 instead of 1.1.0.

• Related issue: STAC 1.1.0 link IRI validation radiantearth/stac-spec#1360

[fmigneault]

See # pragma: has-torch for conditional resolution of 3.10 coverage.

mlm/pyproject.toml

Line 420 in 27bc757

has-torch = "is_installed('torch')"