-
candle-nn
Minimalist ML framework
-
fundsp
Audio processing and synthesis library
-
tch
Rust wrappers for the PyTorch C++ api (libtorch)
-
scirs2-special
Special functions module for SciRS2 (scirs2-special)
-
scirs2-text
Text processing module for SciRS2 (scirs2-text)
-
scirs2-neural
Neural network building blocks module for SciRS2 (scirs2-neural) - Minimal Version
-
onnx-ir
pure Rust library for parsing ONNX models into an intermediate representation that can be used to generate code for various ML/DL frameworks
-
trueno
High-performance SIMD compute library with GPU support for matrix operations
-
cjc-cli
Command-line interface for the CJC compiler and runtime
-
ruvector-attention
Attention mechanisms for ruvector - geometric, graph, and sparse attention
-
candle-flash-attn
Flash attention layer for the candle ML framework
-
muxer
Deterministic, multi-objective bandit-style routing primitives (windowed stats + Pareto selection)
-
irithyll
Streaming ML in Rust -- gradient boosted trees, neural architectures (TTT/KAN/MoE/Mamba/SNN), AutoML, kernel methods, and composable pipelines
-
ruvllm
LLM serving runtime with Ruvector integration - Paged attention, KV cache, and SONA learning
-
pmetal-metal
Metal GPU compute kernels for PMetal - FlashAttention and optimized ML primitives
-
flodl
— a flow-graph deep learning framework built on libtorch
-
g_math
Multi-domain fixed-point arithmetic with geometric extension: Lie groups, manifolds, ODE solvers, tensors, fiber bundles — zero-float, 0 ULP transcendentals
-
ruvector-mincut-gated-transformer
Ultra low latency transformer inference with mincut-gated coherence control
-
cjc-runtime
Runtime library: values, builtins, tensors, COW buffers
-
scirs2-autograd
Automatic differentiation module for SciRS2 (scirs2-autograd)
-
axonml-nn
Neural network modules for Axonml ML framework
-
piper-plus-cli
CLI tool for Piper Plus neural text-to-speech synthesis
-
rmlx-cli
Command-line interface for RMLX
-
ruvector-math
Advanced mathematics for next-gen vector search: Optimal Transport, Information Geometry, Product Manifolds
-
torsh-nn
Neural network modules for ToRSh with PyTorch-compatible API
-
ruvector-attn-mincut
Min-cut gating attention operator: dynamic graph-based alternative to softmax attention
-
tribev2
TRIBE v2 — multimodal fMRI brain encoding model inference in Rust
-
rustyml
A high-performance machine learning & deep learning library in pure Rust, offering ML algorithms and neural network support
-
web-rwkv
RWKV language model in pure WebGPU
-
axonml-tensor
N-dimensional tensor operations for the Axonml ML framework
-
computer-says-no
Local embedding service for text classification using ONNX models
-
numr
High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/WebGPU)
-
morok-tensor
High-level lazy tensor API for the Morok ML compiler
-
ruvector-coherence
Coherence measurement proxies for comparing attention mechanisms
-
rank-refine
SIMD-accelerated MaxSim, cosine, diversity (MMR/DPP) for vector search and RAG pipelines
-
tl-lang
A differentiable programming language with tensor support for machine learning
-
pc-rl-core
Predictive Coding Actor-Critic reinforcement learning framework — pure Rust, zero ML dependencies
-
bitnet-core
Core BitNet implementation with fundamental data structures and algorithms
-
oxiwhisper
Pure Rust Whisper inference engine with minimal dependencies
-
smelt-ml
A machine learning framework for Rust, inspired by mlr3
-
cjc-ad
Automatic differentiation: forward-mode dual numbers and reverse-mode tape
-
native_neural_network
Lib no_std Rust for native neural network (.rnn)
-
large
Rust LLM inference implementation
-
kn-cuda-eval
A CUDA executor for neural network graphs
-
whisper-apr
WASM-first automatic speech recognition engine implementing OpenAI Whisper
-
ruvector-consciousness
SOTA consciousness metrics: IIT Φ computation, causal emergence, effective information with SIMD acceleration and sublinear approximations
-
peregrine-ml
A from-scratch deep learning library in Rust — tensors, autograd, NEON SIMD, Metal GPU, and working models
-
weirwood
Privacy-preserving XGBoost inference via Fully Homomorphic Encryption
-
runnx
A minimal, verifiable ONNX runtime implementation in Rust
-
xsf
Rust implementations and bindings for scipy.special functions
-
amari-fusion
Fusion system for combining algebraic structures
-
gbdt
Gradient Boosting Regression Tree in Rust programming language
-
attnres
Attention Residuals (MoonshotAI/Kimi) implementation in Rust using burn
-
amari-tropical
Tropical (max-plus) algebra implementation
-
jax-rs
JAX in Rust - A complete machine learning framework with WebGPU acceleration
-
oximedia-neural
Lightweight neural network inference for media processing — pure Rust tensor ops, conv2d, and pre-defined media models
-
lift-cli
LIFT CLI: Command-line interface — verify, analyse, print, compile, simulate, predict
-
topological-coherence
Toroidal topology primitives for LLM coherence research — Tonnetz geometry, spectral gap, attention masks, Karmonic spectral filter (v7: inference-time bias null, training-time valid)
-
pmetal-cli
Command-line interface for PMetal LLM fine-tuning
-
god-gragh
A graph-based LLM white-box optimization toolbox: topology validation, Lie group orthogonalization, tensor ring compression
-
ruvector-solver-wasm
WASM bindings for RuVector sublinear-time solver
-
neurons
Neural networks from scratch, in Rust
-
lele
Bare-metal machine learning inference library in Rust
-
entrenar
Training & Optimization library with autograd, LoRA, quantization, and model merging
-
rmlx-distributed
Distributed inference (TP/EP) over RDMA for RMLX
-
tflite
Rust bindings for TensorFlow Lite
-
scirs2-wasm
WebAssembly (WASM) bindings for SciRS2 - JavaScript/TypeScript interop for scientific computing
-
kodegen_simd
KODEGEN.ᴀɪ: SIMD optimization library
-
pc_core
Predictive Coding Actor-Critic reinforcement learning framework — pure Rust, zero ML dependencies
-
tang
Math library for physical reality — geometry, spatial algebra, tensor, training, GPU compute, and 3D gaussian splatting
-
scirs2
port of SciPy with AI/ML extensions - Scientific Computing and AI Library (scirs2)
-
lift-core
SSA IR foundation — types, values, operations, blocks, regions, verifier
-
onnx-infer
Pure Rust ONNX inference engine for computer vision models
-
minikanren_1bit_chirho
miniKanren as 1-bit matrix operations - hardware-accelerated logic programming with SIMD, GPU, and FPGA support. Includes Sudoku solver (14μs), N-Queens, constraint propagation.
-
ruvector-profiler
Memory, power, and latency profiling hooks with CSV emitters for benchmarking attention mechanisms
-
ruvector-fpga-transformer
FPGA Transformer backend with deterministic latency, quantization-first design, and coherence gating
-
rust-lstm
A complete LSTM neural network library with training capabilities, multiple optimizers, and peephole variants
-
nabla-ml
A numpy-like library for Rust
-
yscv-model
Neural network layers, training loop, model zoo, and data loading
-
provable-contracts
Papers to Math to Contracts in Code — YAML contract parsing, validation, scaffold generation, and Kani harness codegen for provable Rust kernels
-
turboquant-rs
TurboQuant KV-Cache Quantization — 3-bit compression with zero accuracy loss (Zandieh et al., ICLR 2026)
-
burn-tensor
Tensor library with user-friendly APIs and automatic differentiation support
-
candle-mi
Mechanistic interpretability for language models in Rust, built on candle
-
yscv-kernels
CPU and GPU compute backends with SIMD dispatch and BLAS integration
-
pmetal-mcp
MCP server exposing PMetal ML toolkit
-
runnt
Easy Neural Network for machine learning
-
provable-contracts-cli
CLI for provable-contracts — validate, scaffold, verify, status, audit
-
hypembed
Pure-Rust BERT-compatible text embedding inference for local-first applications
-
wav2vec2-rs
Rust wav2vec2 CTC forced alignment with CPU, wgpu, and CUDA backends
-
train-station
A high-performance, PyTorch inspired, zero dependency Rust machine learning library
-
oxionnx
Pure Rust ONNX inference engine — zero C/C++ dependencies
-
nnl
A high-performance neural network library for Rust with CPU and GPU support
-
yscv-autograd
Dynamic computation graph with tape-based reverse-mode autodiff
-
oxionnx-ops
Operator implementations for OxiONNX — 88+ ONNX operators in Pure Rust
-
ruvector-attention-wasm
High-performance WebAssembly attention mechanisms: Multi-Head, Flash, Hyperbolic, MoE, CGT Sheaf Attention with GPU acceleration for transformers and LLMs
-
minimum_ml
Experimental Machine Learning Library
-
rankit
Learning to Rank: differentiable ranking, LTR losses (RankNet, LambdaRank, ApproxNDCG, ListNet, ListMLE), trainers, and IR evaluation metrics
-
candle-flash-attn-v3
Flash attention v3 layer for the candle ML framework
-
wonnx
ONNX runtime based on wgpu aimed at being a universal GPU runtime, written in Rust
-
bitnet-benchmarks
Comprehensive benchmarking suite for BitNet implementation
-
torsh
A blazingly fast, production-ready deep learning framework written in pure Rust
-
micro-hnsw-wasm
Neuromorphic HNSW vector search with spiking neural networks - 11.8KB WASM for edge AI, ASIC, and embedded systems. Features LIF neurons, STDP learning, winner-take-all, dendritic computation.
-
a3s-power
A3S Power — Privacy-preserving LLM inference for TEE environments
-
rmlx-metal
Metal GPU abstraction layer for RMLX
-
ruvector-attention-unified-wasm
Unified WebAssembly bindings for 18+ attention mechanisms: Neural, DAG, Graph, and Mamba SSM
-
microcnn
A minimal CNN framework in Rust with Quantization
-
tensorrs
Tensors is a lightweight machine learning library in Rust
-
tenflowers
Pure Rust implementation of TensorFlow - A comprehensive deep learning framework
-
instmodel_inference
High-performance neural network inference library with instruction-based execution
-
ronn-core
Core runtime engine for RONN - fundamental tensor operations, data types, and session management
-
pkboost
Shannon-guided gradient boosting for extreme class imbalance with adaptive drift detection. Outperforms XGBoost/LightGBM on imbalanced data.
-
meuron
modular neural network library written in rust for training simple neural networks
-
tenflowers-autograd
Automatic differentiation engine for TenfloweRS
-
instmodel
Instruction-based neural network model training library using Burn
-
nbml
Machine Learning Primitives
-
kuji
Stochastic sampling primitives: Gumbel-Softmax, reservoir sampling, and latent permutations
-
torsh-functional
Functional programming utilities for ToRSh tensors
-
autograd
Tensors and differentiable operations in Rust
-
oxionnx-gpu
GPU compute backend for OxiONNX — wgpu-based MatMul and Conv2D acceleration
-
statoxide
High-performance statistical computing library written in Rust, exposed to Python via PyO3
-
trueno-gpu
Pure Rust PTX generation for NVIDIA CUDA - no LLVM, no nvcc
-
torsh-models
Pre-trained models and model zoo for ToRSh deep learning framework
-
so-utils
High-performance statistical computing library written in Rust, exposed to Python via PyO3
-
tensorlogic-trustformers
Transformer-as-rules: Self-attention and FFN layers as einsum expressions
-
optimiz-rs
High-performance optimization algorithms in Rust with Python bindings
-
webnn-graph
DSL for WebNN graphs
-
minidx
compile-time-sized neural networks
-
torsh-benches
Benchmarking suite for ToRSh
-
trustformers-training
Training infrastructure for TrustformeRS
-
tenflowers-neural
Neural network layers, models and training APIs for TenfloweRS
-
trustformers-wasm
WebAssembly bindings for TrustformeRS transformer library
-
candle-nn-mold
Minimalist ML framework (mold fork with Metal fixes)
-
lumen-nn
A tiny ML framework
-
ferrotorch-nn
Neural network modules for ferrotorch — layers, losses, initialization
-
sensorlm-rs
SensorLM – wearable sensor foundation model in Rust (Burn + WGPU)
-
yscv-onnx
ONNX runtime with 128+ operators, quantization, and graph optimization
-
fynch
Differentiable sorting and ranking: PAVA, Fenchel-Young losses, and O(n log n) FastSoftSort
-
pmetal-distill
Knowledge distillation toolkit for PMetal - KL divergence, hidden state alignment, and offline distillation
-
rust-mlp
Small, from-scratch MLP (dense feed-forward) for Rust: predictable, allocation-free hot path, batched training, and optional fast GEMM backend
-
nanogpt
Nanochat in Rust
-
burn-mlx
MLX backend for Burn deep learning framework - native Apple Silicon GPU acceleration
-
tenebris
A from-scratch machine learning library written in pure Rust
-
ruvector-gnn-wasm
WebAssembly bindings for RuVector GNN with tensor compression and differentiable search
-
micro_cartan_attn
Complete Cartan matrix attention mechanisms with proper Lie algebra structures
-
briny_ai
A tiny & efficient AI inference engine
-
cuttle
A large language model inference engine in Rust
-
axonml-onnx
ONNX import/export support for the Axonml ML framework
-
candle-cudnn-attn
cuDNN attention bindings for candle - flash-attention compatible API
-
tauformer
A Transformer Architecture using arrowspace's taumode
-
logprob
A wrapper around floats to handle log probabilities
-
nsl-errors
Error types and diagnostic reporting for the NeuralScript compiler
-
cubek-attention
CubeK: Attention Kernels
-
turbo-quant
TurboQuant, PolarQuant, and QJL — zero-overhead vector quantization for semantic search and KV cache compression (ICLR 2026)
-
rten-vecmath
SIMD vectorized implementations of various math functions used in ML models
-
ghostflow-ml
Classical ML algorithms for GhostFlow
-
mininn
A minimalist deep learnig crate for rust
-
mlx-core
Safe Rust API for MLX tensors, devices, and core operations
-
bandit
Algorithms in Rust
-
wasmedge_tensorflow_interface
that provides Rust to WebAssembly developers with syntax for using tensorflow functionality
-
minillm
A mini inference engine for running transformer language models
-
nano-consciousness
High-performance consciousness-inspired AI framework with nanosecond scheduling and temporal processing
-
lightgbm-rust
Rust bindings for LightGBM, a gradient boosting framework. Downloads LightGBM binaries at build time for cross-platform compatibility.
-
scivex-nn
Scivex — Neural networks, autograd, layers, optimizers
-
runn
feature-rich, easy-to-use library for building, training, and evaluating feed-forward neural networks in Rust
-
haagenti
Next-generation compression library with LZ4, Zstd, Brotli, and Deflate support
-
frankensearch-fusion
RRF fusion, score blending, and two-tier progressive search for frankensearch
-
ghostflow-autograd
Automatic differentiation for GhostFlow ML framework
-
trueno-explain
PTX/SIMD/wgpu visualization and tracing CLI for Trueno
-
rllama
Pure Rust implementation of LLaMA-family of models, executable
-
cogent
Basic neural network library for classification
-
pure-onnx-ocr
Pure Rust OCR pipeline that runs PaddleOCR DBNet + SVTR ONNX models without C/C++ dependencies
-
jiro_nn
Neural Networks framework with model building & data preprocessing features
-
rapl
that makes numerical scripting with Rust simple and enjoyable
-
cargo-cog
A cargo subcommand to build, run and publish machine learning containers
-
smelte-rs
Efficient inference ML framework written in rust
-
minidx-vis
Visualization crate for minidx, a simple + compile-time neural-network library
-
phynexus
Empowering Intelligent Futures, One Edge at a Time
-
brique-rs
A MLP library made from scratch, using the rust standard lib
-
wonnx-cli
CLI for WONNX. WONNX is an ONNX runtime based on wgpu aimed at being a universal GPU runtime, written in Rust.
-
cld3
Rust binding for Compact Language Detector v3 (CLD3), a neural network model for language identification. the C++ code is from https://github.com/google/cld3
-
microflow
A robust and efficient TinyML inference engine
-
neural_networks_rust
Neural Networks framework with model specification & data preprocessing features
-
gradients
An OpenCL, CUDA and CPU based Deep Learning Library
-
kaffe
Pytorch, but make it rust :)
-
cubecl-attention
CubeCL Attention Kernels Engine
-
fast-neural-network
A heavily parallelized neural network library designed for speed and flexability
-
burn_attention
Flash Attention v3 implementation for Burn framework
-
collenchyma-nn
collenchyma plugin providing Neural Network operations
-
hanzo-nn
Minimalist ML framework
-
newron
train and infer deep learning models
-
ducky-learn
ML rust framework
-
nove_lossfn
lightweight deep learning library wrapped around Candle Tensor
-
rai-models
ML framework with Ergonomic APIs in Rust
-
tf-rs
High level (Pythonic) API for the TensorFlow framework
-
rmlx-macros
Procedural macros for RMLX
-
stonnx
run inference on ONNX models
-
RayBNN_Optimizer
Gradient Descent Optimizers and Genetic Algorithms using GPUs, CPUs, and FPGAs via CUDA, OpenCL, and oneAPI
-
nevermind-neu
Machine learning library with terminal user interface
-
block-graph
Uses the Burn library to provide block level graph neural network structure
-
irithyll-core
Core types, training engine, and inference for irithyll streaming ML — no_std + alloc, histogram binning, Hoeffding trees, SGBT ensembles, drift detection, f32 + int16 packed formats
-
trustformers-core
Core traits and utilities for TrustformeRS
-
neurox
providing core building blocks for numerical computing and machine learning, including tensor operations, model abstractions, and future GPU acceleration support
-
linearalgebra
Test project to learn Rust and implement a small library for linear algebra
-
dendritic-metrics
Metrics package for dendritic
-
alumina
An Experimental Deep Learning Library
-
logsumexp
Numerically stable computation of log-sum-exp using a 1-pass algorithm
-
rmlx-core
Core GPU operations and kernel registry for RMLX
-
minidx-core
Core crate for minidx, a simple + compile-time neural-network library
-
blasphemy
Inspired by Keras, powered by BLAS. Construct neural networks with one line of code per layer. That's BLASphemy.
-
puffpastry
basic rust neural network framework
-
tangram_linear
Tangram makes it easy for programmers to train, deploy, and monitor machine learning models
-
learntk
machine learning toolkit, written in Rust
-
yarnn
Neural Network framework
-
provable-contracts-macros
Proc macros for compile-time contract enforcement — #[contract] attribute
-
coaster-nn
coaster plugin providing Neural Network operations
-
tsuga
An early-stage machine learning library for building neural networks in Rust
-
hal-ml
HAL: a machine learning library that is able to run on Nvidia, OpenCL or CPU BLAS based compute backends. It currently provides stackable classical neural networks, RNN's and soon to be LSTM's…
-
dumbnet
a [no_std] neural network library
-
neuro
A deep learning library that runs on the GPU
-
pyrus-nn
Micro Neural Network framework
-
gpt-model
Pure-Rust inference wrapper for GPT-2 large language models
-
mnist_read
Reads generic data and label files in the MNIST file format
-
gumbel-top-bucket
A bucket data structure that is sampled using the Gambel-Top trick, which allows for O(1) sampling from a softmax-like distribution
-
ffnn
A very simple feedfoward neural network library
-
diffusion_rs_common
Blazingly fast inference of diffusion models
-
mushin_derive
Compile-time creation of neural networks
-
sticker
Neural sequence labeler
Try searching with DuckDuckGo or on crates.io.