model_handler

Classes

ModelHandler

ModelHandler(model: ModelT, subgraph: SubGraphT, operators: list[AotOperator], config: ConvertArgs)

Functions

compute_values

compute_values(mem_alloc: dict[int, int], total_mem: int) -> dict[str, str]

Computes the values needed for generating the model source code.

Parameters:

• mem_alloc

  (dict[int, int]) –

  Memory allocation for each tensor.

• total_mem

  (int) –

  Total memory size.

Returns:

• dict ( dict[str, str] ) –

  A dictionary containing the computed values for code generation.

generate_source_code

generate_source_code(mem_alloc: dict[int, int], total_mem: int, save_path: Path)

Generate the source code for the main inference model.

Parameters:

• mem_alloc

  (dict[int, int]) –

  Memory allocation for each tensor.

• total_mem

  (int) –

  Total memory size.

• save_path

  (Path) –

  Path to save the generated code.

Functions