Installation

With Python >= 3.10:

pip install sparsepixels

This pulls in everything you need to build and train sparse models: TensorFlow, Keras 3, HGQ2 (the quantization backend), and our built-in monitoring and plotting tools.

Converting to HLS

Deploying a trained model to FPGA firmware additionally needs hls4ml. Support for sparsepixels is merged into hls4ml (PR #1468) but is not in a tagged release yet, so install hls4ml from the main branch:

pip install "git+https://github.com/fastmachinelearning/hls4ml.git"

Once a release including it is out, plain pip install hls4ml will be enough.

See HLS Conversion for details.