LaurentMazare/tch-rs

Rust bindings for the C++ api of PyTorch.

What it solves

tch-rs provides Rust bindings for the PyTorch C++ API (libtorch), allowing developers to use PyTorch's powerful tensor operations, automatic differentiation, and neural network modules directly within the Rust programming language. This enables the creation of high-performance AI models and applications while leveraging Rust's safety and concurrency features.

How it works

The project provides thin wrappers around the C++ libtorch library. It allows users to interact with PyTorch tensors and neural network components through a Rust interface that stays close to the original C++ API. The crate can link against a system-wide libtorch installation, a Python PyTorch installation, or download a pre-built binary of libtorch automatically.

Who it’s for

Developers who want to use PyTorch's deep learning capabilities in Rust, whether for training models from scratch, running inference on pre-trained weights (including support for SafeTensors), or deploying models via TorchScript JIT.

Highlights

  • PyTorch Integration: Direct bindings to libtorch for tensor operations and automatic differentiation.
  • Neural Network API: Tools to build architectures using nn::seq and nn::VarStore for parameter management.
  • Flexible Linking: Supports system-wide, Python-based, or automatic binary downloads of libtorch.
  • Weight Loading: Ability to import pre-trained weights from PyTorch using the SafeTensors format.
  • Broad Application Examples: Includes implementations for MNIST, ResNet, character-level RNNs, and a simplified GPT.

Related

  • Project
    PyO3/rust-numpy

    Rust bindings for the NumPy C-API that allow developers to integrate high-performance Rust numerical computing with Python's NumPy arrays.

  • Project
    hasktorch/hasktorch

    A Haskell library for tensors and neural networks that leverages the core C++ libraries shared by PyTorch.

  • Project
    pykeio/ort

    A Rust interface for hardware-accelerated inference and training of ONNX machine learning models, enabling deployment across various hardware and frameworks.

  • Project
    tracel-ai/burn

    Burn is a comprehensive deep learning framework and tensor library in Rust that unifies training and inference in a single codebase for seamless deployment across diverse hardware.

  • Project
    pytorch/pytorch

    A Python-based machine learning framework that provides GPU-accelerated tensor computation and a dynamic autograd system for building deep neural networks.