NF2 v0.4.0

NF2 v0.4.0

This release introduces the public v0.4 configuration schema, refreshed documentation, updated example workflows, and improved packaging for PyPI and conda.

Highlights

• Added the public v0.4 YAML-style configuration schema.
• Updated Cartesian, spherical, series, benchmark, export, and metrics examples.
• Refreshed documentation, including generated CLI, dataset, configuration, and export/metrics references.
• Added and updated Colab and notebook workflows with clearer defaults and release-ready settings.
• Improved series data loading, reload behavior, lazy loading, and step activation.
• Improved export and metrics CLI behavior for Cartesian and spherical outputs.
• Added WandB logging improvements for training figures.
• Updated dependency constraints, including PyTorch/torchvision compatibility and Python 3.11/3.12 support.
• Added regression tests for release-scale sampler defaults and WandB image logging.

Compatibility Notes

• NF2 v0.4 uses the new configuration schema.
• Legacy keys such as base_path, work_directory, loss, data.type, and lambda are rejected with migration-oriented errors.
• Python support is >=3.11,<3.13.

Verification

Release checks passed locally:

• Test suite: 28 passed
• Source and wheel build: passed
• twine check: passed
• Sphinx documentation build: passed
• Conda recipe render: passed
• CLI smoke checks: passed
• Example configuration normalization: passed