Skip to content

Enabling Real-Time Inference of Temporal Convolution Networks on Low-Power MCUs with Stream-Oriented Automatic Transformation

Notifications You must be signed in to change notification settings

ahmad-mirsalari/StreamEase

Repository files navigation

StreamEase

A python library that automates the conversion of Temporal Convolutional Network (TCN) models to a streaming format without affecting model accuracy, significantly reducing deployment time and computational resources.

Reference

This library is based on the research outlined in the following paper:

Setup

Clone the Repository:

git clone [email protected]:ahmad-mirsalari/TCN_Pre_release.git
cd StreamEase

Install Dependencies:

You can install the required dependencies by running:

pip install -r requirements.txt

Set Up the tool:

Run the setup_env.sh script to add the project root to your PYTHONPATH. This will ensure that Python can locate the streamease package.

chmod +x setup_env.sh
./setup_env.sh

You will be prompted to add the PYTHONPATH to your .bashrc to make it permanent for future sessions.

Repository Organization

  • streamease: Main package containing modules for streaming inference, buffering, and quantization.
  • examples: Contains example scripts demonstrating how to use the toolkit.
  • utils: Contains helper functions to facilitate running the streaming network, with support for ONNX Runtime and GreenWaves Technologies' nntool for optimized model execution.
  • setup_env.sh: Script for setting up dependencies and environment variables.
  • requirements.txt: List of required dependencies

Roadmap

  • Extend the library by adding additional networks in different applications

License

StreamEase is released under Apache 2.0, see the LICENSE file in the root of this repository for details.

Acknowledgements

This work was supported by the APROPOS project (g.a. no. 956090), founded by the European Union’s Horizon 2020 research and innovation program.

Contributors

🚀 Contact Me

About

Enabling Real-Time Inference of Temporal Convolution Networks on Low-Power MCUs with Stream-Oriented Automatic Transformation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published