# nsx-cmsis-nn

<!-- nsx:facts -->
Ambiq-optimized CMSIS-NN neural network kernels. Fork of ARM CMSIS-NN with MVE and DSP optimizations for Apollo SoCs.

- Type: Algorithm
- Version: 7.47.0
- Project: ns-cmsis-nn
- Revision: v7.47.0
- Source: https://github.com/AmbiqAI/ns-cmsis-nn/tree/v7.47.0
- Capabilities: cmsis_nn, neural_network_kernels, quantized_inference, float_inference, dsp_acceleration, mve_acceleration
<!-- /nsx:facts -->

## Add it to an app

```bash
nsx module add nsx-cmsis-nn
```

## Capabilities

- cmsis_nn
- neural_network_kernels
- quantized_inference
- float_inference
- dsp_acceleration
- mve_acceleration

## Use cases

- quantized neural network inference on Cortex-M
- optimized convolution and fully-connected kernels
- int8/int16/int4 quantized model support
- fp32/fp16 floating-point model support

## Requires

- [`nsx-soc-hal`](https://ambiqai.github.io/neuralspotx/modules/nsx-soc-hal/)

## Compatibility

:::note[How compatibility is declared]
The boards, SoCs and toolchains listed here are the ones the module manifest declares.
That is the module author's statement of intent, not evidence from hardware. Confirm on your
own target before you depend on it.
:::

| Kind | Declared |
| --- | --- |
| Boards | * |
| SoCs | * |
| Toolchains | arm-none-eabi-gcc, armclang, atfe |

## Agent keywords

`cmsis-nn`, `ns-cmsis-nn`, `kernels`, `quantization`, `inference`, `dsp`, `mve`, `fp32`, `fp16`, `float`

## Source

| Field | Value |
| --- | --- |
| Type | Algorithm |
| Version | 7.47.0 |
| Project | ns-cmsis-nn |
| Revision | `v7.47.0` |
| Manifest | `modules/ns-cmsis-nn/nsx/nsx-module.yaml` |
| Repository | [ns-cmsis-nn](https://github.com/AmbiqAI/ns-cmsis-nn/tree/v7.47.0) |

Back to the [module catalog](https://ambiqai.github.io/neuralspotx/modules/catalog/).