Enum anira_engine#

enum anira_engine#

Inference engine, one of the two independent backend axes. Values after ANIRA_ENGINE_TFLITE and below ANIRA_ENGINE_CUSTOM are reserved for later built-in engines, which are appended after TFLITE and never renumber CUSTOM. The pair rule, wherever an engine is named (a model entry, the default engine, anira_backend_id, anira_plan_info, anira_edge_info, anira_provider_option_set, anira_processor_ctx): the engine_id is non-NULL if and only if the engine is ANIRA_ENGINE_CUSTOM. A built-in engine is its value with a NULL id; a custom engine (anira_custom_engine_create, added to a pipeline with anira_pipeline_add_engine) is ANIRA_ENGINE_CUSTOM with its reverse-URI id (a processor reads it with anira_processor_engine); ANIRA_ENGINE_NONE names no engine and takes no id. A setter refuses any other combination with ANIRA_ERROR_INVALID_ARGUMENT.

Values:

enumerator ANIRA_ENGINE_NONE#

No engine; as a default engine it means plan 0 (the first plan of the table).

enumerator ANIRA_ENGINE_ONNXRUNTIME#

ONNX Runtime (v2 ONNX, JSON “onnxruntime”).

enumerator ANIRA_ENGINE_EXECUTORCH#

ExecuTorch (JSON “executorch”).

enumerator ANIRA_ENGINE_LITERT#

LiteRT (JSON “litert”).

enumerator ANIRA_ENGINE_LIBTORCH#

LibTorch (JSON “libtorch”).

enumerator ANIRA_ENGINE_TFLITE#

TensorFlow Lite, legacy C API (JSON “tflite”).

enumerator ANIRA_ENGINE_CUSTOM#

A custom engine, named by the engine_id beside it (JSON: the id itself, “com.example.engine”).

enumerator ANIRA_ENGINE_FORCE32#