It’s here:
Python
src/winml/modelkit/export/config.py
batch_size: int = 1 # Static batch for QNN compatibility
Then auto-resolution uses that default explicitly:
Python
src/winml/modelkit/export/config.py
export_config = _resolve_export_config_from_specs(
...
batch_size=WinMLExportConfig().batch_size,
**(shape_config or {}),
)
It’s here:
Python
src/winml/modelkit/export/config.py
batch_size: int = 1 # Static batch for QNN compatibility
Then auto-resolution uses that default explicitly:
Python
src/winml/modelkit/export/config.py
export_config = _resolve_export_config_from_specs(
...
batch_size=WinMLExportConfig().batch_size,
**(shape_config or {}),
)