OptimizationArgs

OptimizationArgs()

Model-wide optimization settings: the goal, the approximation gate and the knobs.

Every knob defaults to auto, which a static table resolves from goal and allow_approximate; operators[].optimization overrides a knob per operator type or id. The resolved choice of every operator is written to <prefix>_plan.json. Knob fields mirror the knob registry, helia_aot.optimization.KNOBS.

goal: OptimizationGoal = Field(default=OptimizationGoal.BALANCED, description="What auto knob values favor: latency, balanced (today's defaults), size or accuracy")

What auto favors. balanced keeps today’s defaults.

allow_approximate: bool = Field(default=False, description="Allow auto knob values that change numerics relative to the knob's default (e.g. fast accumulation)")

Allow auto to choose knob values that change numerics relative to the knob’s default. An explicit knob value is used as given.

accumulation: Accumulation = Field(default=Accumulation.AUTO, description='FP16 fully connected and depthwise accumulation: precise (four-tap FP32 folding on eligible packed fully connected layers, otherwise standard; the default depthwise entry), fast (legacy fully connected routing; the FP16-lane depthwise entry; approximate where alternatives apply) or auto; FP16 convolution packs at every setting')

FP16 accumulation for kernels that offer a choice: precise, fast or auto.

kernel: KernelChoice = Field(default=KernelChoice.AUTO, description='Int8 convolution, depthwise, fully connected and bilinear resize entry: specialized (shape-specialized ns-cmsis-nn direct entries where they apply), generic (the general entries, fewer kernels linked) or auto; both are exact')

Int8 convolution, depthwise and fully connected entry: specialized, generic or auto.

budget: dict[str, float] | None = Field(None, description='Constraints for the optimizer, e.g. accuracy_drop or tcm_bytes', json_schema_extra=stability('reserved'))

Reserved for the optimizer’s constraints (for example accuracy_drop in points or tcm_bytes); setting it is an error.

plan: Path | None = Field(None, description='Path to an optimization plan to read as input: its schedule section is applied, and every other section must describe the conversion the configuration produces', json_schema_extra=stability('experimental'))

An optimization plan to read as input. Its schedule section is applied; every other section must describe the conversion the configuration produces, by the comparison rule of helia_aot.aot.plan_input. A path in a configuration file resolves relative to that file; the command line flag resolves relative to the working directory.