# helia_edge.models.mlperf_tiny_params

Typed parameters of the MLPerf Tiny family; importable without Keras.

## helia_edge.models.mlperf_tiny_params.MlperfTinyParams

`class` · `python`

```python
MlperfTinyParams()
```

Select a faithful fixed architecture; initialization and export belong to callers.

Source: `helia_edge/models/mlperf_tiny_params.py:8`

### helia_edge.models.mlperf_tiny_params.MlperfTinyParams.model_config

`attribute` · `python`

```python
model_config = ConfigDict(frozen=True, extra='forbid')
```

Source: `helia_edge/models/mlperf_tiny_params.py:17`

### helia_edge.models.mlperf_tiny_params.MlperfTinyParams.family

`attribute` · `python`

```python
family: Literal['mlperf_tiny'] = 'mlperf_tiny'
```

Model family

Source: `helia_edge/models/mlperf_tiny_params.py:19`

### helia_edge.models.mlperf_tiny_params.MlperfTinyParams.architecture

`attribute` · `python`

```python
architecture: Literal['kws', 'vww', 'resnet', 'ad']
```

DS-CNN keyword spotting, MobileNetV1 visual wake
words, ResNet-8 image classification or the dense anomaly-detection autoencoder

Source: `helia_edge/models/mlperf_tiny_params.py:20`

### helia_edge.models.mlperf_tiny_params.MlperfTinyParams.input_shape

`attribute` · `python`

```python
input_shape: tuple[int, ...]
```

The architecture's fixed input shape, without the batch axis.

Source: `helia_edge/models/mlperf_tiny_params.py:23`