Welcome to SparseTIR’s documentation!

SparseTIR is a tensor-level compiler for sparse/irregular operators in Deep Learning. The design goal of SparseTIR is to provide a general programming abstraction that can cover both sparse and irregular (e.g. Ragged Tensors) workloads in Deep Learning including Graph Neural Networks, Sparse Transformers, Sparse Convolutions, Network Pruning, etc. while generating high-performance code on heterogeneous hardware.

Getting Started

• Install SparseTIR
• The Blitz Course to SparseTIR

Tutorials

• Sparse Dense Matrix Multiplication (SpMM)
• Sampled Dense Dense Matrix Multiplication (SDDMM)
• Optimizing Unstructured SpMM with Tensor Cores
• Block Sparse Operators
• Relational Graph Convolution
• 3D Sparse Convolution

System Overview

• Why designing another compiler abstraction for Sparse Deep Learning
• System Overview

Misc

• Frequently Asked Questions

API Reference

• tvm.sparse