mlperf_tiny

Faithful scale-1 MLPerf Tiny architectures with newly initialized weights.

Pinned topology and licensing: https://ambiqai.github.io/helia-edge/examples/mlperf-tiny/ and models/licenses/. Adapted from MLCommons Tiny (Apache-2.0); anomaly detector originally Copyright (c) 2020 Hitachi, Ltd. (MIT). Changes: Keras 3 API, fixed captured shapes/configurations, explicit layer names, no training or weight downloads. These constructors do not provide official trained weights or accuracy claims.

Machine-readable model

• buildfunctionBuild a faithful MLPerf Tiny architecture without loading or converting weights.

function

build

Python

Build a faithful MLPerf Tiny architecture without loading or converting weights.

helia_edge/models/mlperf_tiny.py:187

build(
    params: MlperfTinyParams,
    input_shape: tuple[int | None, ...] | None = None,
    *,
    batch_size: int | None = None,
    name: str | None = None,
) -> keras.Model

Build a faithful MLPerf Tiny architecture without loading or converting weights.

Parameters

Parameters of build

Name	Type	Default	Description
params	MlperfTinyParams	Required	The architecture: ``kws``, ``vww``, ``resnet`` or ``ad``.
input_shape	tuple[int | None, ...] | None	None	None, or the architecture's fixed input shape.
batch_size	int | None	None	Static batch size; None for a dynamic batch.
name	str | None	None	Model name; the family when None.

Returns

Returns of build

Type	Description
keras.Model	keras.Model: The model, named ``mlperf_tiny`` unless ``name`` is given.