Sparsity Python Packages

torchao

PyTorch native quantization and sparsity for training and inference

nncf

Neural Network Compression Framework for enhanced OpenVINO™ inference

tensorflow-model-optimization

A toolkit to optimize ML models for deployment for Keras and TensorFlow, including quantization and pruning.

neural-compressor

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

tf-model-optimization-nightly

A suite of tools that users, both novice and advanced can use to optimize machine learning models for deployment and execution.

sparseml

Libraries for applying sparsification recipes to neural networks with a few lines of code, enabling faster and smaller models

sporco

Sparse Optimisation Research Code

neural-compressor-pt

Repository of Intel® Neural Compressor

asdex

Automatic Sparse Differentiation in JAX.

sconce

E2E AutoML Model Compression Package

neural-compressor-tf

Repository of Intel® Neural Compressor

fasterai

FasterAI: Prune and Distill your models with FastAI and PyTorch

neural-speed

Repository of Intel® Intel Extension for Transformers

neural-compressor-full

Repository of Intel® Neural Compressor

neural-solution

Repository of Intel® Neural Compressor

neural-insights

Repository of Intel® Neural Compressor

lpot

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

magni

A package for AFM image reconstruction and compressed sensing in general

randomly

A Library for Denoising Single-Cell Data with Random Matrix Theory

tensorflow-engram

tensorflow-engram: A Python package for Engram Neural Networks, adding biologically-inspired Hebbian memory and engram layers to TensorFlow/Keras models, supporting memory traces, plasticity, attention, and sparsity for neural sequence learning.

ilit

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

gslr

An algorithm for classification from a graph-sparse support

veloce

WIP. Veloce is a low-code Ray-based parallelization library that makes machine learning computation novel, efficient, and heterogeneous.

neural-compressor-3x-pt

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime