A curated collection of Rust projects related to neural networks, designed to complement "Are We Learning Yet."
Python
73
20 commits
updated Aug 22, 2026
A curated collection of Rust projects related to neural networks, designed to complement "Are We Learning Yet". I plan to submit libraries to "Are We Learning Yet" that are included in this project but currently missing from it.
This repository aims to provide a comprehensive overview of the Rust neural network ecosystem to identify the currently missing foundational infrastructure. Thus, it focuses more on the features, backends, and main dependencies of the Rust-based neural network projects. Note that the features, backends, and main dependencies are based on my own search and understanding, and may be incomplete or even incorrect.
| Name | Description | Features | Backends | Main Deps | Last Commit Time |
|---|---|---|---|---|---|
| candle | Minimalist ML framework for Rust | Like PyTorch, Training, Various Models | CPU, CUDA, CUDA NCCL, WASM | gemm, intel-mkl-src, cudarc, metal, accelerate-src | 2026-06-26 |
| burn | Burn is a new comprehensive dynamic Deep Learning Framework built using Rust with extreme flexibility, compute efficiency and portability as its primary goals. | Various backends, Kernel Fusion, Training, Various Models, ONNX | WGPU, Candle, Torch, Ndarray, Remote | matrixmultiply, blas-src, libm, openblas-src, ndarray, candle-core, cubecl, cudarc, tch | 2026-06-26 |
| dfdx | Deep learning in Rust, with shape checked tensors and neural networks | Compile-time Checking | CPU, CUDA, WGPU | gemm, cudarc, wgpu | 2024-01-25 |
| luminal | Search-first deep learning compiler and inference runtime written in Rust | Static Computation Graph, Small Primitive Op Set, Search-based Compilation, Kernel Fusion, Megakernel Compilation, PyTorch torch.compile Backend | CPU, CUDA, Metal | cudarc, metal | 2026-08-22 |
| autograph | A machine learning library for Rust | GPGPU kernels implemented with krnl | CPU, Vulkan | krnl, ndarray | 2024-08-19 |
| unda | General purpose machine learning crate | Compile to XLA | XLA | xla-rs | 2024-06-19 |
| custos | A minimal OpenCL, CUDA, Vulkan and host CPU array manipulation engine / framework | Array Manipulation, AutoDiff, Lazy Execution | CPU, OpenCL, CUDA, Vulkan, NNAPI | min-cl, libm, ash, naga, nnapi | 2025-08-21 |
| zyx | Tensor library for machine learning | Lazy Execution, AutoDiff | CUDA, OpenCL, WGPU | wgpu, vulkano, manual bindings to CUDA, OpenCL, and HSA | 2026-06-26 |
| zenu | A Deep Learning Framework Written in Rust | Training, AutoDiff | CPU, CUDA | cblas, openblas-src, manual binding to CUDA | 2024-12-30 |
| maidenx | A lightweight and fast AI framework in Rust focused on simplicity and performance | Educational focus, Mirror PyTorch's arch | CPU, CUDA | - | 2025-10-19 |
| ort | Fast ML inference & training for ONNX models in Rust | ONNX, Various Backends | CUDA, TensorRT, OpenVINO, oneDNN, DirectML, QNN, CoreML, ACL, TVM, CANN, etc. | ort-sys: Unsafe Rust bindings for ONNX Runtime 1.20 | 2026-06-12 |
| tract | Tiny, no-nonsense, self-contained, Tensorflow and ONNX inference | ONNX, Tensorflow | CPU, Metal | accelerate-src, blis-src, cblas, metal, ndarray, openblas-src, tensorflow, tflitec | 2026-06-27 |
| Kyanite | A neural network inference library, written in Rust | ONNX, Graph IR | CPU, CUDA | manual binding to CUDA | 2026-03-15 |
| rten | ONNX neural network inference engine | ONNX | CPU | rayon | 2026-06-27 |
| oxide-rs | High-performance LLM inference engine inspired by llama.cpp. | LLM Inference, Quantization | CPU | llama.cpp | 2026-03-16 |
| mistral.rs | Blazingly fast LLM inference | LLM inference, safetensors, Quantization | CPU, CUDA, Metal | mkl, candle, metal, accelerate | 2026-06-27 |
| PegaInfer | High-performance LLM inference and serving engine written in Rust with custom CUDA kernels | LLM Serving, Continuous Batching, KV Cache, Prefix Cache, CUDA Graphs, OpenAI-compatible API, Multi-GPU | CUDA | Rust, CUDA | 2026-08-22 |
| hipfire | High-performance LLM inference engine for AMD consumer GPUs written in Rust and HIP | LLM Inference, Quantization, Multi-GPU, Expert Parallelism, Custom GPU Kernels | AMD HIP / ROCm, RDNA | Rust, HIP | 2026-08-22 |
| PMetal | Rust-native machine learning platform for Apple Silicon | LLM Inference, LoRA, QLoRA, DPO, GRPO, Quantization, Training, Serving | Metal, Apple Neural Engine | Rust, Metal | 2026-08-22 |
| rustane | Rust-native training and inference stack targeting Apple Neural Engine and Metal | Neural Network Training, Inference, Apple Neural Engine, Metal GPU | ANE, Metal | Rust, Metal | 2026-08-22 |
| InfiniLM | A handwriting transformer model project developed from YdrMaster/llama2.rs | LLM Inference, Multiple backends supported | CPU, CUDA, OpenCL, Ascend, etc. | operators | 2026-06-26 |
| operators | Multi-hardware support operator library | Multi-hardware | CPU, CUDA, OpenCL, Ascend, Cambricon | clrt, infinirt, cuda-driver | 2025-02-19 |
| crabml | a fast cross platform AI inference engine using Rust and WebGPU | LLM Inference, mmap, Quantization | CPU, WGPU | vulkano, wgpu | 2025-01-04 |
| diffusion-rs | Blazingly fast inference of diffusion models | Diffusion, Quantization, DDUF, Offloading | CPU, CUDA, Metal, etc. | cudarc, intel-mkl-src, accelerate-sr, metal, gemm | 2025-04-01 |
| mmnn | rust-based bash-cli for Neural Network propagation/backpropagation | bash-cli, json-config | - | - | 2025-06-28 |
| Name | Description | Features | Backends | Main Deps | Last Commit Time |
|---|---|---|---|---|---|
| linfa | A Rust machine learning framework | like scikit-learn | CPU | ndarray, sprs | 2026-05-30 |
| jams-rs | Rust based model serving solution for popular machine learning frameworks | Model Store, PyTorch & Tenosrflow Models, Tree Models | PyTorch, TensorFlow, Catboost, LightGBM | tensorflow, tch, catboost-rs, lgbm, ndarray | 2025-05-30 |
| bullet | Specialised ML Library | Domain-specific, NNUE-style networks, chess engines | cudarc | - | 2026-06-22 |
| redstone-ml | High-performance Machine Learning, Dynamic Auto-Differentiation and Tensor Algebra crate for Rust | AutoDiff, NdArrray | CPU, GPU | - | 2025-06-08 |
| RustyML | A high-performance machine learning library in pure Rust, offering statistical utilities, ML algorithms and neural networks. | - | CPU | ndarray, rayon | 2026-06-25 |
| Aprender | Production-oriented machine learning framework written in Rust | Classical ML, Neural Networks, Training, Inference, Quantization, Model Serialization, Profiling | CPU, WGPU, CUDA | ndarray, trueno | 2026-08-22 |
| mni-ml | Machine learning framework with a high-level TypeScript API and performance-critical Rust backend | Tensors, AutoDiff, Neural Networks, Training, Optimizers | CPU, CUDA, WebGPU | Rust native backend | 2026-08-22 |
| tenferro-rs | A Rust-native tensor & autodiff stack for scientific computing | opt-in autodiff, scientific computing | CPU, GPU | faer, cblas, lapack, cudarc, cubecl | 2026-07-28 |
| Name | Description | Features | Backends | Main Deps | Last Commit Time |
|---|---|---|---|---|---|
| cubecl | Multi-platform high-performance compute language and compiler infrastructure for Rust | Kernels in Rust, CubeIR, Comptime Specialization, Autotuning, Kernel Fusion, Tensor Core Acceleration | CUDA, HIP, Metal, SPIR-V, WGSL, CPU | cudarc, cubecl-hip-sys, cubecl-spirv, wgpu | 2026-08-22 |
| krnl | Safe, portable, high performance compute (GPGPU) kernels | Kernels in Rust | Vulkan | rsprv, vulkano, ash | 2025-10-20 |
| EnzymeAD | A rust fork to work towards Enzyme integration | AutoDiff on LLVM | LLVM | Enzyme | 2024-11-25 |
| Rust-CUDA | Ecosystem of libraries and tools for writing and executing fast GPU code fully in Rust | Compiling Rust to PTX | CUDA | Binding | 2026-04-29 |
| Rust-GPU | Making Rust a first-class language and ecosystem for GPU shaders | Compiling Rust to SPIR-V | SPIR-V | Binding | 2026-06-23 |
| cudarc | Safe rust wrapper around CUDA toolkit | Mixture of safe and sys APIs | CUDA | Binding | 2026-06-19 |
| async-cuda | Asynchronous CUDA for Rust | Async, NPP | CUDA | Binding | 2026-06-02 |
| async-tensorrt | Asynchronous TensorRT for Rust | Async, TensorRT | CUDA | async-cuda, Binding | 2026-06-02 |
| cuda-rs | Rust wrapper for CUDA Driver (libcuda.so) and Runtime (libcudart.so) APIs | - | CUDA | Binding | 2024-01-01 |
| tensorrt-rs | Rust wrapper to NVIDIA TensorRT | TensorRT | CUDA | cuda-rs, Binding | 2024-01-03 |
| oxicuda | OxiCUDA replaces the entire NVIDIA CUDA Toolkit software stack with type-safe, memory-safe Rust code | CUDA driver | CUDA | - | 2026-06-25 |
| cuda-oxide | An experimental Rust-to-CUDA compiler that lets you write (SIMT) GPU kernels in safe(ish), idiomatic Rust | Rust to PTX | CUDA | - | 2026-06-27 |
| Rurix | GPU programming language and toolchain written in Rust | CUDA first, Windows first | CUDA | - | 2026-06-25 |
| std::offload / gpu_offload | Experimental native GPU offloading support in Rust/rustc. Paper: GPU Offload in Rust: Portable, Safe, and Fast | Rust Host/Device Offload, Ownership-aware Data Movement, LLVM Offload, Portable GPU Kernels | NVIDIA, AMD, Intel GPUs | rustc, LLVM Offload | 2026-08-22 |
| tile-rs | Safe Rust bindings for Huawei Ascend NPU (CANN framework) | Kernels in Rust tile DSL, lower to MLIR then 15 backends | 15 backends | - | 2026-06-29 |
| Paper | Related Project | Topic | Year |
|---|---|---|---|
| Fearless Concurrency on the GPU | cutile-rs | Safe tile-based GPU programming in Rust, ownership, data-race freedom, high-performance GPU kernels | 2026 |
| GPU Offload in Rust: Portable, Safe, and Fast | rustc / std::offload | Native multi-vendor GPU offloading from Rust through LLVM, ownership-aware host/device data movement | 2026 |
| What Irregularity Costs: CUDA C++, Rust, and Triton on a Hash-Blocked GPU Workload | cuda-oxide | CUDA C++ vs Rust vs Triton for irregular GPU workloads, atomics and data-dependent execution | 2026 |
| LithOS: An Operating System for Efficient Machine Learning on GPUs | LithOS | Rust-based GPU operating system, ML scheduling, GPU resource management and kernel execution | 2025 |
A curated collection of Rust projects related to neural networks, designed to complement "Are We Learning Yet."
Python
73
20 commits
updated Aug 22, 2026
A curated collection of Rust projects related to neural networks, designed to complement "Are We Learning Yet". I plan to submit libraries to "Are We Learning Yet" that are included in this project but currently missing from it.
This repository aims to provide a comprehensive overview of the Rust neural network ecosystem to identify the currently missing foundational infrastructure. Thus, it focuses more on the features, backends, and main dependencies of the Rust-based neural network projects. Note that the features, backends, and main dependencies are based on my own search and understanding, and may be incomplete or even incorrect.
| Name | Description | Features | Backends | Main Deps | Last Commit Time |
|---|---|---|---|---|---|
| candle | Minimalist ML framework for Rust | Like PyTorch, Training, Various Models | CPU, CUDA, CUDA NCCL, WASM | gemm, intel-mkl-src, cudarc, metal, accelerate-src | 2026-06-26 |
| burn | Burn is a new comprehensive dynamic Deep Learning Framework built using Rust with extreme flexibility, compute efficiency and portability as its primary goals. | Various backends, Kernel Fusion, Training, Various Models, ONNX | WGPU, Candle, Torch, Ndarray, Remote | matrixmultiply, blas-src, libm, openblas-src, ndarray, candle-core, cubecl, cudarc, tch | 2026-06-26 |
| dfdx | Deep learning in Rust, with shape checked tensors and neural networks | Compile-time Checking | CPU, CUDA, WGPU | gemm, cudarc, wgpu | 2024-01-25 |
| luminal | Search-first deep learning compiler and inference runtime written in Rust | Static Computation Graph, Small Primitive Op Set, Search-based Compilation, Kernel Fusion, Megakernel Compilation, PyTorch torch.compile Backend | CPU, CUDA, Metal | cudarc, metal | 2026-08-22 |
| autograph | A machine learning library for Rust | GPGPU kernels implemented with krnl | CPU, Vulkan | krnl, ndarray | 2024-08-19 |
| unda | General purpose machine learning crate | Compile to XLA | XLA | xla-rs | 2024-06-19 |
| custos | A minimal OpenCL, CUDA, Vulkan and host CPU array manipulation engine / framework | Array Manipulation, AutoDiff, Lazy Execution | CPU, OpenCL, CUDA, Vulkan, NNAPI | min-cl, libm, ash, naga, nnapi | 2025-08-21 |
| zyx | Tensor library for machine learning | Lazy Execution, AutoDiff | CUDA, OpenCL, WGPU | wgpu, vulkano, manual bindings to CUDA, OpenCL, and HSA | 2026-06-26 |
| zenu | A Deep Learning Framework Written in Rust | Training, AutoDiff | CPU, CUDA | cblas, openblas-src, manual binding to CUDA | 2024-12-30 |
| maidenx | A lightweight and fast AI framework in Rust focused on simplicity and performance | Educational focus, Mirror PyTorch's arch | CPU, CUDA | - | 2025-10-19 |
| ort | Fast ML inference & training for ONNX models in Rust | ONNX, Various Backends | CUDA, TensorRT, OpenVINO, oneDNN, DirectML, QNN, CoreML, ACL, TVM, CANN, etc. | ort-sys: Unsafe Rust bindings for ONNX Runtime 1.20 | 2026-06-12 |
| tract | Tiny, no-nonsense, self-contained, Tensorflow and ONNX inference | ONNX, Tensorflow | CPU, Metal | accelerate-src, blis-src, cblas, metal, ndarray, openblas-src, tensorflow, tflitec | 2026-06-27 |
| Kyanite | A neural network inference library, written in Rust | ONNX, Graph IR | CPU, CUDA | manual binding to CUDA | 2026-03-15 |
| rten | ONNX neural network inference engine | ONNX | CPU | rayon | 2026-06-27 |
| oxide-rs | High-performance LLM inference engine inspired by llama.cpp. | LLM Inference, Quantization | CPU | llama.cpp | 2026-03-16 |
| mistral.rs | Blazingly fast LLM inference | LLM inference, safetensors, Quantization | CPU, CUDA, Metal | mkl, candle, metal, accelerate | 2026-06-27 |
| PegaInfer | High-performance LLM inference and serving engine written in Rust with custom CUDA kernels | LLM Serving, Continuous Batching, KV Cache, Prefix Cache, CUDA Graphs, OpenAI-compatible API, Multi-GPU | CUDA | Rust, CUDA | 2026-08-22 |
| hipfire | High-performance LLM inference engine for AMD consumer GPUs written in Rust and HIP | LLM Inference, Quantization, Multi-GPU, Expert Parallelism, Custom GPU Kernels | AMD HIP / ROCm, RDNA | Rust, HIP | 2026-08-22 |
| PMetal | Rust-native machine learning platform for Apple Silicon | LLM Inference, LoRA, QLoRA, DPO, GRPO, Quantization, Training, Serving | Metal, Apple Neural Engine | Rust, Metal | 2026-08-22 |
| rustane | Rust-native training and inference stack targeting Apple Neural Engine and Metal | Neural Network Training, Inference, Apple Neural Engine, Metal GPU | ANE, Metal | Rust, Metal | 2026-08-22 |
| InfiniLM | A handwriting transformer model project developed from YdrMaster/llama2.rs | LLM Inference, Multiple backends supported | CPU, CUDA, OpenCL, Ascend, etc. | operators | 2026-06-26 |
| operators | Multi-hardware support operator library | Multi-hardware | CPU, CUDA, OpenCL, Ascend, Cambricon | clrt, infinirt, cuda-driver | 2025-02-19 |
| crabml | a fast cross platform AI inference engine using Rust and WebGPU | LLM Inference, mmap, Quantization | CPU, WGPU | vulkano, wgpu | 2025-01-04 |
| diffusion-rs | Blazingly fast inference of diffusion models | Diffusion, Quantization, DDUF, Offloading | CPU, CUDA, Metal, etc. | cudarc, intel-mkl-src, accelerate-sr, metal, gemm | 2025-04-01 |
| mmnn | rust-based bash-cli for Neural Network propagation/backpropagation | bash-cli, json-config | - | - | 2025-06-28 |
| Name | Description | Features | Backends | Main Deps | Last Commit Time |
|---|---|---|---|---|---|
| linfa | A Rust machine learning framework | like scikit-learn | CPU | ndarray, sprs | 2026-05-30 |
| jams-rs | Rust based model serving solution for popular machine learning frameworks | Model Store, PyTorch & Tenosrflow Models, Tree Models | PyTorch, TensorFlow, Catboost, LightGBM | tensorflow, tch, catboost-rs, lgbm, ndarray | 2025-05-30 |
| bullet | Specialised ML Library | Domain-specific, NNUE-style networks, chess engines | cudarc | - | 2026-06-22 |
| redstone-ml | High-performance Machine Learning, Dynamic Auto-Differentiation and Tensor Algebra crate for Rust | AutoDiff, NdArrray | CPU, GPU | - | 2025-06-08 |
| RustyML | A high-performance machine learning library in pure Rust, offering statistical utilities, ML algorithms and neural networks. | - | CPU | ndarray, rayon | 2026-06-25 |
| Aprender | Production-oriented machine learning framework written in Rust | Classical ML, Neural Networks, Training, Inference, Quantization, Model Serialization, Profiling | CPU, WGPU, CUDA | ndarray, trueno | 2026-08-22 |
| mni-ml | Machine learning framework with a high-level TypeScript API and performance-critical Rust backend | Tensors, AutoDiff, Neural Networks, Training, Optimizers | CPU, CUDA, WebGPU | Rust native backend | 2026-08-22 |
| tenferro-rs | A Rust-native tensor & autodiff stack for scientific computing | opt-in autodiff, scientific computing | CPU, GPU | faer, cblas, lapack, cudarc, cubecl | 2026-07-28 |
| Name | Description | Features | Backends | Main Deps | Last Commit Time |
|---|---|---|---|---|---|
| cubecl | Multi-platform high-performance compute language and compiler infrastructure for Rust | Kernels in Rust, CubeIR, Comptime Specialization, Autotuning, Kernel Fusion, Tensor Core Acceleration | CUDA, HIP, Metal, SPIR-V, WGSL, CPU | cudarc, cubecl-hip-sys, cubecl-spirv, wgpu | 2026-08-22 |
| krnl | Safe, portable, high performance compute (GPGPU) kernels | Kernels in Rust | Vulkan | rsprv, vulkano, ash | 2025-10-20 |
| EnzymeAD | A rust fork to work towards Enzyme integration | AutoDiff on LLVM | LLVM | Enzyme | 2024-11-25 |
| Rust-CUDA | Ecosystem of libraries and tools for writing and executing fast GPU code fully in Rust | Compiling Rust to PTX | CUDA | Binding | 2026-04-29 |
| Rust-GPU | Making Rust a first-class language and ecosystem for GPU shaders | Compiling Rust to SPIR-V | SPIR-V | Binding | 2026-06-23 |
| cudarc | Safe rust wrapper around CUDA toolkit | Mixture of safe and sys APIs | CUDA | Binding | 2026-06-19 |
| async-cuda | Asynchronous CUDA for Rust | Async, NPP | CUDA | Binding | 2026-06-02 |
| async-tensorrt | Asynchronous TensorRT for Rust | Async, TensorRT | CUDA | async-cuda, Binding | 2026-06-02 |
| cuda-rs | Rust wrapper for CUDA Driver (libcuda.so) and Runtime (libcudart.so) APIs | - | CUDA | Binding | 2024-01-01 |
| tensorrt-rs | Rust wrapper to NVIDIA TensorRT | TensorRT | CUDA | cuda-rs, Binding | 2024-01-03 |
| oxicuda | OxiCUDA replaces the entire NVIDIA CUDA Toolkit software stack with type-safe, memory-safe Rust code | CUDA driver | CUDA | - | 2026-06-25 |
| cuda-oxide | An experimental Rust-to-CUDA compiler that lets you write (SIMT) GPU kernels in safe(ish), idiomatic Rust | Rust to PTX | CUDA | - | 2026-06-27 |
| Rurix | GPU programming language and toolchain written in Rust | CUDA first, Windows first | CUDA | - | 2026-06-25 |
| std::offload / gpu_offload | Experimental native GPU offloading support in Rust/rustc. Paper: GPU Offload in Rust: Portable, Safe, and Fast | Rust Host/Device Offload, Ownership-aware Data Movement, LLVM Offload, Portable GPU Kernels | NVIDIA, AMD, Intel GPUs | rustc, LLVM Offload | 2026-08-22 |
| tile-rs | Safe Rust bindings for Huawei Ascend NPU (CANN framework) | Kernels in Rust tile DSL, lower to MLIR then 15 backends | 15 backends | - | 2026-06-29 |
| Paper | Related Project | Topic | Year |
|---|---|---|---|
| Fearless Concurrency on the GPU | cutile-rs | Safe tile-based GPU programming in Rust, ownership, data-race freedom, high-performance GPU kernels | 2026 |
| GPU Offload in Rust: Portable, Safe, and Fast | rustc / std::offload | Native multi-vendor GPU offloading from Rust through LLVM, ownership-aware host/device data movement | 2026 |
| What Irregularity Costs: CUDA C++, Rust, and Triton on a Hash-Blocked GPU Workload | cuda-oxide | CUDA C++ vs Rust vs Triton for irregular GPU workloads, atomics and data-dependent execution | 2026 |
| LithOS: An Operating System for Efficient Machine Learning on GPUs | LithOS | Rust-based GPU operating system, ML scheduling, GPU resource management and kernel execution | 2025 |
