femtocrux

femtocrux

Compiler API for the Femotsense Sparse Processing Unit (SPU).

This package drives compilation of machine learning models to SPU binaries, as well as simulation of those binaries on SPU hardware.

The package itself is a thin Python wrapper communicating to the actual compiler / simulator via gRPC.

Supported ML model representations:

• Femtosense Quantized IR (FQIR)

Installation

You will need to install femtocrux and femtodriver to compile models to run on the SPU and to run simulations to get power metrics.

You will need to have python 3.10 and Docker installed. Install Docker by following these instructions.

femtocrux is available on PyPI via

pip install femtocrux

femtodriver is also available on PyPI via

pip install femtodriver

Usage

Please refer to the documentation in the e2e walkthrough for a full example of how to create a model in pytorch, use femtodriver to compile program files to run on the SPU, and for how to run simulations of the model to get power estimates.

If you want more detailed documentation of all the options you can use to compile and simulate with please refer to the femtodriver documentation.