Metadata-Version: 2.1
Name: neuralspot-edge
Version: 0.2.2
Summary: NSE is a Keras add-on targeting edge devices.
Author: Ambiq AI
Author-email: support.aitg@ambiq.com
Requires-Python: >=3.10,<3.13
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Dist: boto3 (>=1.34.151,<2.0.0)
Requires-Dist: h5py (>=3.10.0,<4.0.0)
Requires-Dist: keras (>=3.0.4,<4.0.0)
Requires-Dist: matplotlib (>=3.9.0,<4.0.0)
Requires-Dist: pandas (>=2.2.2,<3.0.0)
Requires-Dist: plotly (>=5.22.0,<6.0.0)
Requires-Dist: pydantic (>=2.6.1,<3.0.0)
Requires-Dist: requests (>=2.31.0,<3.0.0)
Requires-Dist: scikit-learn (>=1.5.1,<2.0.0)
Requires-Dist: seaborn (>=0.13.2,<0.14.0)
Requires-Dist: tensorflow (>=2.16.1,<3.0.0)
Requires-Dist: tqdm (>=4.66.4,<5.0.0)
Description-Content-Type: text/markdown

<p align="center">
  <a href="https://github.com/AmbiqAI/neuralspot-edge"><img src="./docs/assets/nse-banner.png" alt="NSE"></a>
</p>

**Documentation**: <a href="https://ambiqai.github.io/neuralspot-edge" target="_blank">https://ambiqai.github.io/neuralspot-edge</a>

**Source Code**: <a href="https://github.com/AmbiqAI/neuralspot-edge" target="_blank">https://github.com/AmbiqAI/neuralspot-edge</a>

---

# NeuralSPOT Edge (NSE)

<img style="float: left; margin: 0px 15px 15px 0px;" src="./docs/assets/nse-logo-128.png" width="96px" />

neuralSPOT Edge (**NSE**) is [Keras 3](https://keras.io) add-on focused on training and deploying models on resource-constrained, edge devices. NSE relies heavily on [Keras 3](https://keras.io) leveraging it's multi-backend support and customizable architecture. This package provides a variety of additional models, layers, optimizers, quantizers, and other components to help users train and deploy models for edge devices.

<br style="clear:both" />


- **[Getting Started](usage/index.md)**: Learn how to install and use NSE
- **[API Documentation](api/neuralspot_edge)**: Explore the API
- **[Usage Examples](examples/index.md)**: See examples of NSE in action
- **[Explore Guides](guides/index.md)**: View in-depth guides on using NSE

## Main Features

* [**Callbacks**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/callbacks): Training callbacks
* [**Converters**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/converters): Converters for exporting models
* [**Interpreters**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/interpreters): Inference engine interpreters (e.g. TFLite)
* [**Layers**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/layers): Custom layers including `tf.data.Dataset` preprocessing layers
* [**Losses**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/losses): Additional losses such as SimCLRLoss
* [**Metrics**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/metrics): Custom metrics such as SNR
* [**Models**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/models): Highly parameterized 1D/2D model architectures
* [**Optimizers**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/optimizers): Additional optimizers
* [**Plotting**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/plotting): Plotting routines
* [**Quantizers**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/quantizers): Quantization techniques
* [**Trainers**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/trainers): Custom trainers such as SSL contrastive learning
* [**Utils**](https://ambiqai.github.io/neuralspot-edge/api/neuralspot_edge/utils): Utility functions

## Problems **NSE** looks to solve

### Compatability issues between frameworks and inference engines

- [x] By leveraging Keras 3, entire workflows can be run using a variety of backends using a consistent front-end API. This allows selecting a backend that plays nicely with a specific inference engine without rewriting the entire model.

### SOTA models dont scale down well and come in limited configurations

- [x] By providing highly parameterized model architectures based on SOTA models, users can easily scale down models to fit their needs.

### Limited 1D time-series models

- [x] Most included models in NSE provide both 1D and 2D versions. The package also contains time-series specific models.

### Limited support for quantization, pruning, and other model optimization techniques

- [x] NSE provides a variety of quantization and pruning techniques to optimize models for edge deployment.

