Contents Menu Expand Light mode Dark mode Auto light/dark, in light mode Auto light/dark, in dark mode Skip to content
Torch Lattice
Torch Lattice

Getting started

  • Getting started
    • Installation
    • Quickstart
    • Workflow

Reference

  • Concepts
    • Sparse tensor model
    • Convolution semantics
    • Artifact contract
    • Migrating from TorchSparse
  • Tooling
    • Benchmarks
    • Conformance
    • CUDA CI and release build

API reference

  • API reference
    • Core APIs
      • Top-level package
      • Sparse tensor
      • Backend flags
    • Neural network modules
      • NN top-level exports
      • Convolution modules
      • Pooling modules
      • Activation modules
      • Normalization modules
      • BEV modules
      • Crop modules
      • NN utilities
    • Functional APIs
      • Functional top-level exports
      • Functional convolution
      • Convolution configuration
      • Functional pooling
      • Coordinate hashing
      • Coordinate query
      • Sparse relations
      • Voxelization
      • Dense conversion
      • Functional activation
      • Functional crop
      • Sparse count
    • Operator helpers
      • Sparse operators
      • Quantization helpers
      • Tuning helpers
    • Artifact APIs
      • Artifact IO
      • Artifact builder
      • FX lowering
    • Tooling APIs
      • Benchmark tooling
      • Conformance tooling

Project notes

  • Stability policy
  • Compatibility notes
  • Troubleshooting
Back to top
View this page
Edit this page

ConceptsΒΆ

These pages describe the stable sparse semantics used by Torch Lattice and the artifact contract shared with MLX Lattice.

  • Sparse tensor model
  • Convolution semantics
  • Artifact contract
  • Migrating from TorchSparse
Next
Sparse tensor model
Previous
Workflow
Copyright © 2026, Z.Y. Lin
Made with Sphinx and @pradyunsg's Furo