pySpatialML v0.5.2

pySpatialML v0.5.2

This release brings pipeline authoring, validation, and execution into closer alignment with the SpatialML/OpenMR pipeline specification.

Highlights

• Standardized canonical operator names across the CLI, converters, validators, examples, and runtime.
• Replaced generic operator arguments with operand-specific CLI arguments such as --src, --dst, --operand0, and --result.
• Added support for dynamic and named model/JavaScript tensor bindings.
• Centralized operator contracts so pipeline authoring and validation use consistent input and output definitions.
• Strengthened validation for:
  • Tensor descriptors and required placeholders.
  • Operator signatures, operand counts, and tensor types.
  • Model input and output bindings.
  • Canonical tensor reference objects.
  • Package-relative asset and model paths.
  • Invalid absolute paths and directory traversal.
• Improved native XR pipeline loading, model I/O handling, tensor readback, and device execution.
• Refreshed the bundled XR and Spatial device-runner APKs.
• Added XR and Spatial all-operator pipeline fixtures.
• Expanded regression coverage across CLI authoring, conversion, package validation, native loading, and device runners.

Compatibility Note

Existing scripts may need to be updated if they use short operator names or generic --input and --output arguments. Pipeline JSON should use canonical
operator names, numeric tensor type and usage values, canonical tensor references, and package-relative paths.