MODULE REFERENCE

onnx

anomx.models.onnx

Inference from portable ONNX model artifacts.

ONNXModel

ONNXModel(path: str | Path | bytes, *, providers: list[str] | None = None)

Run a published graph without importing or unpickling Python model code.

Signature parameters
Parameter	Type	Default
`path`	`str \| Path \| bytes`	`required`
`providers`keyword only	`list[str] \| None`	`None`

predictmethod

predict(observations: ArrayLike) -> np.ndarray

Predict a batch of (batch, input_window, channels) observations.

Signature parameters
Parameter	Type	Default
`observations`	`ArrayLike`	`required`

Returns np.ndarray

runmethod

run(inputs: Mapping[str, ArrayLike]) -> dict[str, Any]

Run named ONNX tensors and return JSON-compatible named outputs.

Supports numeric, boolean and string tensors, including models with multiple inputs/outputs. The runtime validates tensor dimensions.

Signature parameters
Parameter	Type	Default
`inputs`	`Mapping[str, ArrayLike]`	`required`

Returns dict[str, Any]