Open-source gpu-computing software
544 open-source libraries, tools and applications tagged gpu-computing, ranked by how closely gpu-computing is their primary purpose, then by adoption and maintenance health. Each entry links to a page with sources, use cases and when to choose or avoid it.
544 products, primary matches first, then adoption-weighted; health v2 shown.
| Product | Health v2 | Stars | Maturity |
|---|---|---|---|
| DeepDetect DeepDetect is an open-source deep learning runtime, CLI, and REST server written in C++ for training and inference across images, text, tab… | 95 | 2551 | active |
| yifan123/flow_grpo Flow-GRPO is the official PyTorch implementation of a NeurIPS 2025 paper that trains flow matching models (e.g., SD3.5, FLUX.1, Qwen-Image,… | 52 | 2543 | active |
| google/tunix Tunix is a lightweight JAX-based library for post-training large language models, supporting supervised fine-tuning, preference optimizatio… | 81 | 2490 | active |
| NovaSky-AI/SkyRL SkyRL is a modular full-stack reinforcement learning library for post-training large language models, combining a training framework (skyrl… | 83 | 2436 | active |
| learnhouse/learnhouse LearnHouse is a next-generation open-source learning management system (LMS) for creating, sharing, and selling educational content. It com… | 100 | 2404 | active |
| Olive Olive is Microsoft's AI model optimization toolkit for the ONNX Runtime, automating finetuning, conversion, quantization, and compression o… | 89 | 2399 | active |
| Neroued/ninfer NInfer is a from-scratch C++/CUDA inference engine optimized for maximum single-GPU performance on a narrow, explicitly registered set of Q… | 57 | 2305 | active |
| dstack dstack is an open-source, vendor-agnostic control plane for GPU provisioning and orchestration that works across GPU clouds, Kubernetes, an… | 95 | 2277 | active |
| MFEM MFEM is a lightweight, modular C++ library for finite element discretization of PDEs, supporting arbitrary high-order element spaces, adapt… | 85 | 2258 | stable |
| google-deepmind/mujoco_playground MuJoCo Playground is an open-source Python library of GPU-accelerated robot learning environments built on MuJoCo MJX and MuJoCo Warp. It s… | 74 | 2250 | active |
| MetalPetal/MetalPetal MetalPetal is a GPU-accelerated image and video processing framework built on Apple's Metal API. It provides an image/filter/render pipelin… | 23 | 2188 | active |
| PrimeIntellect-ai/prime-rl prime-rl is a Python framework for large-scale, fully asynchronous reinforcement learning training of language models, built on FSDP2 for t… | 88 | 2164 | active |
| nv-tlabs/vipe ViPE is an open-source video processing engine from NVIDIA that estimates camera intrinsics, camera motion, and dense near-metric depth map… | 77 | 2122 | active |
| patrick-kidger/diffrax Diffrax is a JAX-based library providing numerical differential equation solvers for ODEs, SDEs, and CDEs. It is fully autodifferentiable a… | 80 | 2106 | active |
| marcoslucianops/DeepStream-Yolo A collection of configuration files, parsers, and conversion utilities for running YOLO-family object detection models on NVIDIA DeepStream… | 57 | 2074 | active |
| NVIDIA-NeMo/RL NeMo RL is NVIDIA's open-source post-training library for scaling reinforcement learning methods (GRPO, PPO, DPO, SFT, distillation) on LLM… | 80 | 2073 | active |
| chapel-lang/chapel Chapel is a modern open-source programming language designed for productive parallel computing at scale, with first-class support for task … | 90 | 2047 | active |
| tdrussell/diffusion-pipe A Python training script for fine-tuning diffusion models (image and video generation) using DeepSpeed pipeline parallelism across multiple… | 67 | 2028 | active |
| NUS-HPC-AI-Lab/VideoSys VideoSys is an open-source Python library providing easy and efficient infrastructure for video generation, supporting training, inference,… | 31 | 2025 | active |
| mil-tokyo/webdnn WebDNN is a framework for running deep neural network inference directly in the web browser, accepting ONNX models without Python preproces… | 65 | 2000 | active |
| not-an-aardvark/lucky-commit A Rust CLI tool that amends git commits with whitespace until the commit hash starts with a desired prefix (default '0000000'). It uses GPU… | 37 | 1994 | active |
| Tavris1/ComfyUI-Easy-Install ComfyUI-Easy-Install is a portable one-click installer for ComfyUI that bundles an EZi Desktop application, requiring no manual Python or G… | 84 | 1938 | active |
| AbdBarho/stable-diffusion-webui-docker A Docker Compose setup that runs Stable Diffusion locally with popular web UIs like AUTOMATIC1111, ComfyUI, and InvokeAI. It packages model… | 23 | 7304 | maintenance |
| flexflow/flexflow-train FlexFlow Train is a deep learning framework that accelerates distributed DNN training by automatically searching for efficient parallelizat… | 67 | 1903 | active |
| laugh12321/TensorRT-YOLO A C++/Python deployment toolkit for running YOLO-family models (YOLOv3 through YOLO26) on NVIDIA GPUs using TensorRT, with custom plugins, … | 71 | 1893 | active |
| nazdridoy/kokoro-tts A Python CLI text-to-speech tool built on the Kokoro-82M model that converts text, EPUB, PDF, and TXT inputs into natural-sounding speech w… | 82 | 1889 | active |
| NVIDIA-AI-IOT/Lidar_AI_Solution NVIDIA's collection of GPU-accelerated Lidar AI inference solutions for autonomous driving, including optimized implementations of PointPil… | 69 | 1885 | active |
| koide3/glim GLIM is a versatile and extensible point cloud-based 3D localization and mapping (SLAM) framework written in C++. It performs direct multi-… | 76 | 1881 | active |
| nndeploy/nndeploy nndeploy is an easy-to-use, high-performance AI deployment framework written in C++ with Python bindings. It provides a visual drag-and-dro… | 82 | 1880 | active |
| NVlabs/stylegan3 Official PyTorch implementation of StyleGAN3 (Alias-Free GANs), a state-of-the-art generative adversarial network for high-fidelity image s… | 32 | 6945 | maintenance |
| beam-cloud/beta9 Beam (beta9) is an open-source serverless runtime for AI workloads, providing GPU inference endpoints, isolated sandboxes for running untru… | 91 | 1827 | active |
| triple-mu/YOLOv8-TensorRT A library for running YOLOv8 inference accelerated with NVIDIA TensorRT, supporting detection, segmentation, pose estimation, oriented boun… | 72 | 1812 | active |
| 2swap/swaptube SwapTube is a C++ framework for programmatically rendering YouTube videos, built on FFMPEG with custom graphics code above the encoding lay… | 75 | 1787 | active |
| NVIDIA/FasterTransformer NVIDIA's highly optimized C++/CUDA library for fast inference of Transformer-based models such as BERT, GPT, and encoder-decoder models, wi… | 23 | 6450 | maintenance |
| tenstorrent/tt-metal TT-Metal is Tenstorrent's open-source software stack containing TT-NN, a Python and C++ neural network operator library, and TT-Metalium, a… | 99 | 1716 | active |
| BindsNET/bindsnet BindsNET is a Python package for simulating spiking neural networks (SNNs) built on PyTorch tensor functionality, running on CPUs or GPUs. … | 81 | 1704 | active |
| tkarras/progressive_growing_of_gans Official TensorFlow implementation of the ICLR 2018 NVIDIA paper 'Progressive Growing of GANs', which trains generators and discriminators … | 32 | 6176 | maintenance |
| GoMLX GoMLX is an accelerated machine learning and math framework for Go, comparable to PyTorch/JAX/TensorFlow. It offers differentiable operator… | 96 | 1651 | active |
| NVlabs/sionna Sionna is an open-source, GPU-accelerated, differentiable Python library from NVIDIA for research on communication systems. It comprises Si… | 91 | 1615 | active |
| gorgonia/gorgonia Gorgonia is a Go library for machine learning that lets you define and evaluate mathematical equations over multidimensional arrays using a… | 23 | 5927 | maintenance |
| Tencent-Hunyuan/HY-WorldPlay HY-WorldPlay (HY-World 1.5) is Tencent Hunyuan's open-source framework for interactive 3D world modeling, generating explorable 3D scenes f… | 52 | 1604 | active |
| alibaba/Pai-Megatron-Patch Pai-Megatron-Patch is an open-source deep learning training toolkit from Alibaba Cloud for large-scale training and inference of LLMs and V… | 50 | 1599 | active |
| Tencent-Hunyuan/HunyuanWorld-Voyager HunyuanWorld-Voyager is a video diffusion framework from Tencent Hunyuan that generates world-consistent RGBD video and 3D point-cloud sequ… | 49 | 1599 | active |
| pytorch/FBGEMM FBGEMM is a collection of highly optimized low-precision matrix multiplication and convolution kernels for server-side deep learning infere… | 91 | 1595 | active |
| Xilinx/brevitas Brevitas is a PyTorch library for neural network quantization supporting both post-training quantization (PTQ) and quantization-aware train… | 93 | 1582 | active |
| Soul-AILab/SoulX-FlashTalk SoulX-FlashTalk is a 14B audio-driven talking avatar model that streams infinite real-time video from a reference image and audio, achievin… | 56 | 1526 | active |
| gpu-io gpu-io is a TypeScript WebGL library for composing GPU-accelerated computing workflows in the browser. It handles WebGL state management, s… | 32 | 1486 | active |
| Lightning-AI/lightning-thunder Lightning Thunder is a source-to-source deep learning compiler for PyTorch that optimizes models for training and inference. It provides a … | 70 | 1475 | active |
| FeiYull/TensorRT-Alpha A C++/CUDA library providing TensorRT-accelerated deployment for 30+ popular computer vision models including YOLOv3-v8, YOLOv8-Pose/Seg/Cl… | 77 | 1462 | active |
| tensorflow/tpu A collection of reference models and tools for training machine learning models on Google Cloud TPUs, maintained as a public mirror by the … | 68 | 5278 | maintenance |
| deepseek-ai/EPLB EPLB is DeepSeek's open-source Expert Parallelism Load Balancer for Mixture-of-Experts models. It computes balanced expert replication and … | 23 | 1443 | active |
| trilinos/Trilinos Trilinos is a collection of C++ libraries and an object-oriented software framework for solving large-scale, complex multi-physics engineer… | 98 | 1439 | stable |
| mratsim/Arraymancer Arraymancer is a fast, ergonomic N-dimensional tensor (ndarray) library written in Nim, inspired by NumPy and PyTorch. It provides CPU, CUD… | 57 | 1407 | active |
| davideberly/GeometricTools The Geometric Tools Engine (GTE) is a C++14 collection of source code for computing in mathematics, geometry, graphics, image analysis, and… | 77 | 1386 | active |
| amandaghassaei/OrigamiSimulator A realtime WebGL web application that simulates how any origami crease pattern folds, solving all creases simultaneously via GPU fragment s… | 52 | 1381 | active |
| nomadkaraoke/python-audio-separator A Python package and CLI that separates audio files into stems (vocals, instrumental, drums, bass, etc.) using pre-trained models from Ulti… | 90 | 1378 | active |
| MACE MACE is a Python library implementing fast and accurate machine learning interatomic potentials using higher-order equivariant message pass… | 86 | 1373 | active |
| k2-fsa/k2 k2 is a C++/CUDA library with Python bindings that implements differentiable Finite State Automaton (FSA) and Finite State Transducer (FST)… | 61 | 1368 | active |
| OpenPPL/ppl.nn PPLNN is a high-performance deep-learning inference engine written in C++ that runs ONNX models on x86 CPUs and NVIDIA GPUs, with a dedicat… | 32 | 1368 | active |
| NVIDIA-AI-IOT/torch2trt torch2trt is a Python library that converts PyTorch models to TensorRT engines using the TensorRT Python API, with a simple single-function… | 23 | 4885 | maintenance |
| jonathan-laurent/AlphaZero.jl A generic, simple, and fast Julia implementation of DeepMind's AlphaZero algorithm for training game-playing agents via self-play and MCTS.… | 70 | 1338 | active |
| BytedTsinghua-SIA/CUDA-Agent CUDA-Agent is a large-scale agentic reinforcement learning system from ByteDance Seed and Tsinghua that trains LLMs to generate high-perfor… | 53 | 1338 | active |
| facebookincubator/AITemplate AITemplate is a Python framework that compiles deep neural networks into high-performance CUDA (NVIDIA) or HIP (AMD) C++ code for fast fp16… | 67 | 4730 | maintenance |
| anvaka/fieldplay Field Play is a browser-based WebGL application for exploring and visualizing vector fields by animating thousands of GPU-driven particles.… | 75 | 1328 | active |
| meta-pytorch/segment-anything-fast A fast, batched offline inference-oriented fork of Meta's Segment Anything (SAM) image segmentation model. It applies optimizations like bf… | 42 | 1321 | active |
| derrian-distro/LoRA_Easy_Training_Scripts A PySide6 desktop GUI that wraps Kohya's sd-scripts to simplify training LoRA, LoCon, and other LoRA-type models for Stable Diffusion. It s… | 30 | 1304 | active |
| plaidml/plaidml PlaidML is a portable tensor compiler that enables deep learning on hardware (especially GPUs and embedded devices) not well supported by m… | 10 | 4566 | maintenance |
| ModelCloud/GPTQModel GPTQModel is a production-ready Python toolkit for quantizing (compressing) large language models using GPTQ, AWQ, and related methods, wit… | 92 | 1274 | active |
| fastgs/FastGS FastGS is a general acceleration framework for 3D Gaussian Splatting that trains scenes in roughly 100 seconds using multi-view consistent … | 46 | 1265 | active |
| Ksuriuri/index-tts-vllm A reimplementation of IndexTTS's GPT model inference using vLLM, providing significantly faster text-to-speech generation with a web UI and… | 53 | 1243 | active |
| chengzeyi/Comfy-WaveSpeed A ComfyUI custom node plugin that acts as an all-in-one inference optimization solution for diffusion models, built around First Block Cach… | 63 | 1238 | stable |
| antvis/G G is a flexible 2D/3D rendering engine for visualization, serving as the underlying graphics engine of the AntV charting ecosystem. It adap… | 67 | 1215 | active |
| dendenxu/fast-gaussian-rasterization A drop-in replacement for diff-gaussian-rasterization that renders 3D Gaussian Splatting scenes using a geometry-shader-based GPU pipeline … | 16 | 1215 | active |
| Soul-AILab/SoulX-LiveAct SoulX-LiveAct is the official inference code for a real-time human animation framework that generates lifelike, audio/multimodal-controlled… | 51 | 1193 | active |
| steelbrain/ffmpeg-over-ip A client/server tool that lets applications use GPU-accelerated ffmpeg on a remote machine over a single TCP connection, without GPU passth… | 78 | 1187 | active |
| Linaom1214/TensorRT-For-YOLO-Series A Python and C++ toolkit for running YOLO-series object detection models (YOLOv3 through YOLOv12, YOLOX) with NVIDIA TensorRT, including ON… | 40 | 1158 | active |
| open-gigaai/giga-world-1 GigaWorld-1 is an open-source framework providing training, inference, data processing, checkpoint conversion, and LoRA merge workflows for… | 52 | 1152 | active |
| flagos-ai/FlagGems FlagGems is a high-performance operator library for large language models written in the Triton language, providing backend-neutral GPU ker… | 88 | 1137 | active |
| AutoArk/open-audio-opd An industrial training stack for online policy distillation (OPD) of audio models, distilling compact ASR (and planned TTS) student models … | 49 | 1128 | active |
| Omniverse Kit Sample Suite A template toolkit from NVIDIA for building GPU-accelerated, OpenUSD-based 3D applications with the Omniverse Kit SDK. It provides pre-conf… | 74 | 1106 | active |
| NVlabs/Fast-dLLM NVIDIA's official implementation of Fast-dLLM, a family of training-free and fine-tuning-based acceleration techniques for diffusion-based … | 54 | 1090 | active |
| mlc-ai/web-stable-diffusion A project that compiles and runs Stable Diffusion text-to-image models entirely inside web browsers using WebGPU and WebAssembly, with no s… | 31 | 3724 | maintenance |
| arcee-ai/DistillKit DistillKit is an open-source Python toolkit for knowledge distillation of large language models, supporting both online and offline distill… | 57 | 1076 | active |
| bytedance/SandboxFusion A secure, self-hosted code sandbox service from ByteDance that runs and judges code generated by LLMs across 20+ programming languages via … | 60 | 1072 | active |
| open-gigaai/giga-train GigaTrain is an efficient and scalable Python training framework for large AI models, supporting distributed multi-GPU/multi-node execution… | 59 | 1072 | active |
| cuOpt NVIDIA cuOpt is a GPU-accelerated optimization engine for linear programming, quadratic programming, mixed integer programming, and vehicle… | 83 | 1067 | active |
| LuisaGroup/LuisaCompute LuisaCompute is a high-performance cross-platform computing framework for graphics and beyond, featuring a C++-embedded DSL for GPU kernel … | 77 | 1055 | active |
| NVIDIA/raft NVIDIA RAFT (Reusable Accelerated Functions and Tools) is a C++/Python library of CUDA-accelerated algorithms and primitives for machine le… | 90 | 1043 | active |
| BabitMF/bmf BMF (Babit Multimedia Framework) is a cross-platform, multi-language multimedia and video processing framework developed by ByteDance, offe… | 72 | 1033 | active |
| kaito-project/kaito KAITO is a Kubernetes operator suite that automates LLM inference, fine-tuning, and RAG engine deployment using simplified CRD APIs. It aut… | 93 | 1028 | active |
| fla-org/native-sparse-attention Efficient Triton kernel implementations of Native Sparse Attention (NSA), a hardware-aligned, natively trainable sparse attention mechanism… | 47 | 1023 | active |
| microsoft/Tutel Tutel is Microsoft's optimized Mixture-of-Experts (MoE) library for efficient training and inference of large language models, featuring dy… | 77 | 1019 | active |
| Pjbomb2/TrueTrace-Unity-Pathtracer A high-performance compute shader based path tracer for Unity3D that works without RT cores, using compressed wide BVH for software ray tra… | 75 | 1016 | active |
| Sxela/WarpFusion WarpFusion is a Stable Diffusion-based video-to-video style transfer tool distributed as a Jupyter/Colab notebook. It applies AI animation … | 27 | 1016 | active |
| MoonshotAI/checkpoint-engine Checkpoint-engine is a lightweight Python middleware for updating model weights in-place across LLM inference engines, a critical step in r… | 78 | 1005 | active |
| catalyst-team/catalyst Catalyst is a high-level PyTorch framework for deep learning research and development, focused on reproducibility, rapid experimentation, a… | 60 | 3386 | maintenance |
| XYZ-AI-Lab/axrl AxisRL is an agentic reinforcement learning post-training framework for large language models, built on SGLang for high-throughput rollout … | 53 | 984 | active |
| bytedance/lightseq LightSeq is a high-performance CUDA-based library for training and inference of sequence models like Transformer, BERT, GPT, and BART, with… | 10 | 3292 | maintenance |
| google/brax Brax is a fast, fully differentiable rigidbody physics engine written in JAX, designed for massively parallel simulation on accelerator har… | 84 | 3247 | maintenance |
| tusen-ai/simpledet SimpleDet is a Python framework built on MXNet for object detection and instance recognition. It provides state-of-the-art detection models… | 32 | 3085 | maintenance |