TylerHilbert/PyTorchConference2025_GithubRepos

1[2  {3    "repo_name": "llvm-project",4    "repo_link": "https://github.com/llvm/llvm-project",5    "category": "compiler",6    "github_about_section": "The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.",7    "homepage_link": "http://llvm.org"8  },9  {10    "repo_name": "pytorch",11    "repo_link": "https://github.com/pytorch/pytorch",12    "category": "machine learning framework",13    "github_about_section": "Tensors and Dynamic neural networks in Python with strong GPU acceleration",14    "homepage_link": "https://pytorch.org"15  },16  {17    "repo_name": "kubernetes",18    "repo_link": "https://github.com/kubernetes/kubernetes",19    "category": "container orchestration",20    "github_about_section": "Production-Grade Container Scheduling and Management",21    "homepage_link": "https://kubernetes.io"22  },23  {24    "repo_name": "tensorflow",25    "repo_link": "https://github.com/tensorflow/tensorflow",26    "category": "machine learning framework",27    "github_about_section": "An Open Source Machine Learning Framework for Everyone",28    "homepage_link": "https://tensorflow.org"29  },30  {31    "repo_name": "transformers",32    "repo_link": "https://github.com/huggingface/transformers",33    "category": "multi-purpose library",34    "github_about_section": "Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.",35    "homepage_link": "https://huggingface.co/transformers"36  },37  {38    "repo_name": "spark",39    "repo_link": "https://github.com/apache/spark",40    "category": "data processing",41    "github_about_section": "Apache Spark - A unified analytics engine for large-scale data processing",42    "homepage_link": "https://spark.apache.org"43  },44  {45    "repo_name": "hhvm",46    "repo_link": "https://github.com/facebook/hhvm",47    "category": "virtual machine",48    "github_about_section": "A virtual machine for executing programs written in Hack.",49    "homepage_link": "https://hhvm.com"50  },51  {52    "repo_name": "vllm",53    "repo_link": "https://github.com/vllm-project/vllm",54    "category": "inference engine",55    "github_about_section": "A high-throughput and memory-efficient inference and serving engine for LLMs",56    "homepage_link": "https://docs.vllm.ai"57  },58  {59    "repo_name": "elasticsearch",60    "repo_link": "https://github.com/elastic/elasticsearch",61    "category": "search engine",62    "github_about_section": "Free and Open Source, Distributed, RESTful Search Engine",63    "homepage_link": "https://elastic.co/products/elasticsearch"64  },65  {66    "repo_name": "numpy",67    "repo_link": "https://github.com/numpy/numpy",68    "category": "scientific computing",69    "github_about_section": "The fundamental package for scientific computing with Python.",70    "homepage_link": "https://numpy.org"71  },72  {73    "repo_name": "scipy",74    "repo_link": "https://github.com/scipy/scipy",75    "category": "scientific computing",76    "github_about_section": "SciPy library main repository",77    "homepage_link": "https://scipy.org"78  },79  {80    "repo_name": "llama.cpp",81    "repo_link": "https://github.com/ggml-org/llama.cpp",82    "category": "inference engine",83    "github_about_section": "LLM inference in C/C++",84    "homepage_link": "https://ggml.ai"85  },86  {87    "repo_name": "ray",88    "repo_link": "https://github.com/ray-project/ray",89    "category": "multi-purpose library",90    "github_about_section": "Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.",91    "homepage_link": "https://ray.io"92  },93  {94    "repo_name": "sglang",95    "repo_link": "https://github.com/sgl-project/sglang",96    "category": "inference engine",97    "github_about_section": "SGLang is a high-performance serving framework for large language models and multimodal models.",98    "homepage_link": "https://www.sglang.io"99  },100  {101    "repo_name": "rocm-systems",102    "repo_link": "https://github.com/ROCm/rocm-systems",103    "category": "multi-purpose library",104    "github_about_section": "super repo for rocm systems projects",105    "homepage_link": "https://amd.com/en/products/software/rocm.html"106  },107  {108    "repo_name": "jax",109    "repo_link": "https://github.com/jax-ml/jax",110    "category": "scientific computing",111    "github_about_section": "Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more",112    "homepage_link": "https://docs.jax.dev"113  },114  {115    "repo_name": "onnxruntime",116    "repo_link": "https://github.com/microsoft/onnxruntime",117    "category": "machine learning interoperability",118    "github_about_section": "ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator",119    "homepage_link": "https://onnxruntime.ai"120  },121  {122    "repo_name": "verl",123    "repo_link": "https://github.com/volcengine/verl",124    "category": "reinforcement learning",125    "github_about_section": "verl: Volcano Engine Reinforcement Learning for LLMs",126    "homepage_link": "https://verl.readthedocs.io"127  },128  {129    "repo_name": "ollama",130    "repo_link": "https://github.com/ollama/ollama",131    "category": "inference engine",132    "github_about_section": "Get up and running with OpenAI gpt-oss, DeepSeek-R1, Gemma 3 and other models.",133    "homepage_link": "https://ollama.com"134  },135  {136    "repo_name": "Triton",137    "repo_link": "https://github.com/triton-lang/triton",138    "category": "parallel computing dsl",139    "github_about_section": "A language and compiler for writing highly efficient custom Deep-Learning primitives.",140    "homepage_link": "https://triton-lang.org"141  },142  {143    "repo_name": "executorch",144    "repo_link": "https://github.com/pytorch/executorch",145    "category": "model compiler",146    "github_about_section": "On-device AI across mobile, embedded and edge for PyTorch",147    "homepage_link": "https://executorch.ai"148  },149  {150    "repo_name": "triton-windows",151    "repo_link": "https://github.com/woct0rdho/triton-windows",152    "category": "parallel computing dsl",153    "github_about_section": "Fork of the Triton language and compiler for Windows support and easy installation"154  },155  {156    "repo_name": "trl",157    "repo_link": "https://github.com/huggingface/trl",158    "category": "reinforcement learning",159    "github_about_section": "Train transformer language models with reinforcement learning.",160    "homepage_link": "http://hf.co/docs/trl"161  },162  {163    "repo_name": "goose",164    "repo_link": "https://github.com/block/goose",165    "category": "agent",166    "github_about_section": "an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM",167    "homepage_link": "https://block.github.io/goose"168  },169  {170    "repo_name": "DeepSpeed",171    "repo_link": "https://github.com/deepspeedai/DeepSpeed",172    "category": "training framework",173    "github_about_section": "DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.",174    "homepage_link": "https://deepspeed.ai"175  },176  {177    "repo_name": "numba",178    "repo_link": "https://github.com/numba/numba",179    "category": "compiler",180    "github_about_section": "NumPy aware dynamic Python compiler using LLVM",181    "homepage_link": "https://numba.pydata.org"182  },183  {184    "repo_name": "torchrec",185    "repo_link": "https://github.com/meta-pytorch/torchrec",186    "category": "distributed computing",187    "github_about_section": "Pytorch domain library for recommendation systems",188    "homepage_link": "https://meta-pytorch.org/torchrec/"189  },190  {191    "repo_name": "modular",192    "repo_link": "https://github.com/modular/modular",193    "category": "parallel computing",194    "github_about_section": "The Modular Platform (includes MAX & Mojo)",195    "homepage_link": "https://www.modular.com"196  },197  {198    "repo_name": "accelerate",199    "repo_link": "https://github.com/huggingface/accelerate",200    "category": "training framework",201    "github_about_section": "A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support.",202    "homepage_link": "https://huggingface.co/docs/accelerate"203  },204  {205    "repo_name": "milvus",206    "repo_link": "https://github.com/milvus-io/milvus",207    "category": "vector database",208    "github_about_section": "Milvus is a high-performance, cloud-native vector database built for scalable vector ANN search",209    "homepage_link": "https://milvus.io"210  },211  {212    "repo_name": "onnx",213    "repo_link": "https://github.com/onnx/onnx",214    "category": "machine learning interoperability",215    "github_about_section": "Open standard for machine learning interoperability",216    "homepage_link": "https://onnx.ai"217  },218  {219    "repo_name": "modelcontextprotocol",220    "repo_link": "https://github.com/modelcontextprotocol/modelcontextprotocol",221    "category": "mcp",222    "github_about_section": "Specification and documentation for the Model Context Protocol",223    "homepage_link": "https://modelcontextprotocol.io"224  },225  {226    "repo_name": "ComfyUI",227    "repo_link": "https://github.com/comfyanonymous/ComfyUI",228    "category": "user interface",229    "github_about_section": "The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.",230    "homepage_link": "https://comfy.org"231  },232  {233    "repo_name": "peft",234    "repo_link": "https://github.com/huggingface/peft",235    "category": "fine tuning",236    "github_about_section": "PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.",237    "homepage_link": "https://huggingface.co/docs/peft"238  },239  {240    "repo_name": "flashinfer",241    "repo_link": "https://github.com/flashinfer-ai/flashinfer",242    "category": "gpu kernels",243    "github_about_section": "FlashInfer: Kernel Library for LLM Serving",244    "homepage_link": "https://flashinfer.ai"245  },246  {247    "repo_name": "trainer",248    "repo_link": "https://github.com/kubeflow/trainer",249    "github_about_section": "Distributed AI Model Training and LLM Fine-Tuning on Kubernetes",250    "homepage_link": "https://www.kubeflow.org/docs/components/trainer/"251  },252  {253    "repo_name": "CUTLASS",254    "repo_link": "https://github.com/NVIDIA/cutlass",255    "category": "parallel computing",256    "github_about_section": "CUDA Templates and Python DSLs for High-Performance Linear Algebra",257    "homepage_link": "https://docs.nvidia.com/cutlass/latest/"258  },259  {260    "repo_name": "aiter",261    "repo_link": "https://github.com/ROCm/aiter",262    "category": "gpu kernels",263    "github_about_section": "AI Tensor Engine for ROCm",264    "homepage_link": "https://rocm.blogs.amd.com/software-tools-optimization/aiter-ai-tensor-engine/README.html"265  },266  {267    "repo_name": "wandb",268    "repo_link": "https://github.com/wandb/wandb",269    "category": "ml visualization",270    "github_about_section": "The AI developer platform. Use Weights & Biases to train and fine-tune models, and manage models from experimentation to production.",271    "homepage_link": "https://wandb.ai"272  },273  {274    "repo_name": "TorchAO",275    "repo_link": "https://github.com/pytorch/ao",276    "category": "quantization",277    "github_about_section": "PyTorch native quantization for training and inference",278    "homepage_link": "https://pytorch.org/ao"279  },280  {281    "repo_name": "Mooncake",282    "repo_link": "https://github.com/kvcache-ai/Mooncake",283    "category": "inference",284    "github_about_section": "Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.",285    "homepage_link": "https://kvcache-ai.github.io/Mooncake"286  },287  {288    "repo_name": "ccache",289    "repo_link": "https://github.com/ccache/ccache",290    "category": "compiler",291    "github_about_section": "ccache - a fast compiler cache",292    "homepage_link": "https://ccache.dev"293  },294  {295    "repo_name": "unsloth",296    "repo_link": "https://github.com/unslothai/unsloth",297    "category": "fine tuning",298    "github_about_section": "Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more.",299    "homepage_link": "https://unsloth.ai/docs"300  },301  {302    "repo_name": "torchtitan",303    "repo_link": "https://github.com/pytorch/torchtitan",304    "category": "training framework",305    "github_about_section": "A PyTorch native platform for training generative AI models",306    "homepage_link": "https://arxiv.org/abs/2410.06511"307  },308  {309    "repo_name": "LMCache",310    "repo_link": "https://github.com/LMCache/LMCache",311    "category": "inference",312    "github_about_section": "Supercharge Your LLM with the Fastest KV Cache Layer",313    "homepage_link": "https://lmcache.ai"314  },315  {316    "repo_name": "lapack",317    "repo_link": "https://github.com/Reference-LAPACK/lapack",318    "category": "linear algebra",319    "github_about_section": "LAPACK is a library of Fortran subroutines for solving the most commonly occurring problems in numerical linear algebra.",320    "homepage_link": "https://netlib.org/lapack"321  },322  {323    "repo_name": "FlagGems",324    "repo_link": "https://github.com/flagos-ai/FlagGems",325    "github_about_section": "FlagGems is an operator library for large language models implemented in the Triton Language."326  },327  {328    "repo_name": "ROCm",329    "repo_link": "https://github.com/ROCm/ROCm",330    "category": "multi-purpose library",331    "github_about_section": "AMD ROCm Software - GitHub Home",332    "homepage_link": "https://rocm.docs.amd.com"333  },334  {335    "repo_name": "letta",336    "repo_link": "https://github.com/letta-ai/letta",337    "category": "agent",338    "github_about_section": "Letta is the platform for building stateful agents: open AI with advanced memory that can learn and self-improve over time.",339    "homepage_link": "https://docs.letta.com"340  },341  {342    "repo_name": "Triton Inference Server",343    "repo_link": "https://github.com/triton-inference-server/server",344    "category": "inference server",345    "github_about_section": "The Triton Inference Server provides an optimized cloud and edge inferencing solution.",346    "homepage_link": "https://docs.nvidia.com/deeplearning/triton-inference-server/user-guide/docs/index.html"347  },348  {349    "repo_name": "aws-neuron-sdk",350    "repo_link": "https://github.com/aws-neuron/aws-neuron-sdk",351    "category": "sdk",352    "github_about_section": "Powering AWS purpose-built machine learning chips. Blazing fast and cost effective, natively integrated into PyTorch and TensorFlow and integrated with your favorite AWS services",353    "homepage_link": "https://aws.amazon.com/ai/machine-learning/neuron"354  },355  {356    "repo_name": "Liger-Kernel",357    "repo_link": "https://github.com/linkedin/Liger-Kernel",358    "category": "kernel examples",359    "github_about_section": "Efficient Triton Kernels for LLM Training",360    "homepage_link": "https://openreview.net/pdf?id=36SjAIT42G"361  },362  {363    "repo_name": "tilelang",364    "repo_link": "https://github.com/tile-ai/tilelang",365    "category": "parallel computing dsl",366    "github_about_section": "Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels",367    "homepage_link": "https://tilelang.com"368  },369  {370    "repo_name": "metaflow",371    "repo_link": "https://github.com/Netflix/metaflow",372    "category": "container orchestration",373    "github_about_section": "Build, Manage and Deploy AI/ML Systems",374    "homepage_link": "https://metaflow.org"375  },376  {377    "repo_name": "litgpt",378    "repo_link": "https://github.com/Lightning-AI/litgpt",379    "github_about_section": "20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.",380    "homepage_link": "https://lightning.ai/"381  },382  {383    "repo_name": "monarch",384    "repo_link": "https://github.com/meta-pytorch/monarch",385    "category": "distributed computing",386    "github_about_section": "Monarch is a distributed programming framework for PyTorch based on scalable actor messaging.",387    "homepage_link": "https://meta-pytorch.org/monarch"388  },389  {390    "repo_name": "hipBLASLt",391    "repo_link": "https://github.com/AMD-AGI/hipBLASLt",392    "category": "Basic Linear Algebra Subprograms (BLAS)",393    "github_about_section": "hipBLASLt is a library that provides general matrix-matrix operations with a flexible API and extends functionalities beyond a traditional BLAS library",394    "homepage_link": "https://rocm.docs.amd.com/projects/hipBLASLt"395  },396  {397    "repo_name": "flash-linear-attention",398    "repo_link": "https://github.com/fla-org/flash-linear-attention",399    "category": "gpu kernels",400    "github_about_section": "Efficient implementations of state-of-the-art linear attention models"401  },402  {403    "repo_name": "TensorRT",404    "repo_link": "https://github.com/NVIDIA/TensorRT",405    "category": "inference engine",406    "github_about_section": "NVIDIA TensorRT is an SDK for high-performance deep learning inference on NVIDIA GPUs. This repository contains the open source components of TensorRT.",407    "homepage_link": "https://developer.nvidia.com/tensorrt"408  },409  {410    "repo_name": "AReal",411    "repo_link": "https://github.com/inclusionAI/AReaL",412    "category": "reinforcement learning",413    "github_about_section": "The RL Bridge for LLM-based Agent Applications. Made Simple & Flexible.",414    "homepage_link": "https://www.inclusion-ai.org/AReaL"415  },416  {417    "repo_name": "terminal-bench",418    "repo_link": "https://github.com/laude-institute/terminal-bench",419    "category": "benchmark",420    "github_about_section": "A benchmark for LLMs on complicated tasks in the terminal",421    "homepage_link": "https://tbench.ai"422  },423  {424    "repo_name": "warp",425    "repo_link": "https://github.com/NVIDIA/warp",426    "category": "spatial computing",427    "github_about_section": "A Python framework for accelerated simulation, data generation and spatial computing.",428    "homepage_link": "https://nvidia.github.io/warp"429  },430  {431    "repo_name": "OpenRLHF",432    "repo_link": "https://github.com/OpenRLHF/OpenRLHF",433    "category": "reinforcement learning",434    "github_about_section": "An Easy-to-use, Scalable and High-performance Agentic RL Framework based on Ray (PPO & DAPO & REINFORCE++ & VLM & TIS & vLLM & Ray & Async RL)",435    "homepage_link": "https://openrlhf.readthedocs.io"436  },437  {438    "repo_name": "truss",439    "repo_link": "https://github.com/basetenlabs/truss",440    "category": "inference engine",441    "github_about_section": "The simplest way to serve AI/ML models in production",442    "homepage_link": "https://truss.baseten.co"443  },444  {445    "repo_name": "RLinf",446    "repo_link": "https://github.com/RLinf/RLinf",447    "category": "reinforcement learning",448    "github_about_section": "RLinf: Reinforcement Learning Infrastructure for Embodied and Agentic AI",449    "homepage_link": "https://rlinf.readthedocs.io"450  },451  {452    "repo_name": "nccl",453    "repo_link": "https://github.com/NVIDIA/nccl",454    "category": "distributed computing",455    "github_about_section": "Optimized primitives for collective multi-GPU communication",456    "homepage_link": "https://developer.nvidia.com/nccl"457  },458  {459    "repo_name": "SkyRL",460    "repo_link": "https://github.com/NovaSky-AI/SkyRL",461    "category": "reinforcement learning",462    "github_about_section": "SkyRL: A Modular Full-stack RL Library for LLMs",463    "homepage_link": "https://docs.skyrl.ai/docs"464  },465  {466    "repo_name": "ROLL",467    "repo_link": "https://github.com/alibaba/ROLL",468    "category": "reinforcement learning",469    "github_about_section": "An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models",470    "homepage_link": "https://alibaba.github.io/ROLL/"471  },472  {473    "repo_name": "lightning-thunder",474    "repo_link": "https://github.com/Lightning-AI/lightning-thunder",475    "category": "model compiler",476    "github_about_section": "PyTorch compiler that accelerates training and inference. Get built-in optimizations for performance, memory, parallelism, and easily write your own.",477    "homepage_link": "https://lightning.ai/docs/pytorch/stable"478  },479  {480    "repo_name": "Helion",481    "repo_link": "https://github.com/pytorch/helion",482    "category": "parallel computing dsl",483    "github_about_section": "A Python-embedded DSL that makes it easy to write fast, scalable ML kernels with minimal boilerplate.",484    "homepage_link": "https://helionlang.com"485  },486  {487    "repo_name": "ort",488    "repo_link": "https://github.com/pykeio/ort",489    "category": "machine learning interoperability",490    "github_about_section": "Fast ML inference & training for ONNX models in Rust",491    "homepage_link": "https://ort.pyke.io"492  },493  {494    "repo_name": "deepinv",495    "repo_link": "https://github.com/deepinv/deepinv",496    "github_about_section": "DeepInverse: a PyTorch library for solving imaging inverse problems using deep learning",497    "homepage_link": "https://deepinv.github.io/deepinv/"498  },499  {500    "repo_name": "dstack",501    "repo_link": "https://github.com/dstackai/dstack",502    "category": "container orchestration",503    "github_about_section": "dstack is an open-source control plane for running development, training, and inference jobs on GPUs-across hyperscalers, neoclouds, or on-prem.",504    "homepage_link": "https://dstack.ai"505  },506  {507    "repo_name": "doctr",508    "repo_link": "https://github.com/mindee/doctr",509    "github_about_section": "docTR (Document Text Recognition) - a seamless, high-performing & accessible library for OCR-related tasks powered by Deep Learning.",510    "homepage_link": "https://mindee.github.io/doctr/"511  },512  {513    "repo_name": "SWE-bench",514    "repo_link": "https://github.com/SWE-bench/SWE-bench",515    "category": "benchmark",516    "github_about_section": "SWE-bench: Can Language Models Resolve Real-world Github Issues?",517    "homepage_link": "https://swebench.com"518  },519  {520    "repo_name": "torchchat",521    "repo_link": "https://github.com/pytorch/torchchat",522    "github_about_section": "Run PyTorch LLMs locally on servers, desktop and mobile"523  },524  {525    "repo_name": "mcp-agent",526    "repo_link": "https://github.com/lastmile-ai/mcp-agent",527    "category": "mcp",528    "github_about_section": "Build effective agents using Model Context Protocol and simple workflow patterns"529  },530  {531    "repo_name": "prime-rl",532    "repo_link": "https://github.com/PrimeIntellect-ai/prime-rl",533    "category": "reinforcement learning",534    "github_about_section": "Agentic RL Training at Scale"535  },536  {537    "repo_name": "cuda-python",538    "repo_link": "https://github.com/NVIDIA/cuda-python",539    "category": "middleware",540    "github_about_section": "CUDA Python: Performance meets Productivity",541    "homepage_link": "https://nvidia.github.io/cuda-python"542  },543  {544    "repo_name": "open-instruct",545    "repo_link": "https://github.com/allenai/open-instruct",546    "category": "reinforcement learning",547    "github_about_section": "AllenAI's post-training codebase",548    "homepage_link": "https://allenai.github.io/open-instruct/"549  },550  {551    "repo_name": "openevolve",552    "repo_link": "https://github.com/codelion/openevolve",553    "category": "evolutionary algorithm",554    "github_about_section": "Open-source implementation of AlphaEvolve"555  },556  {557    "repo_name": "litserve",558    "repo_link": "https://github.com/Lightning-AI/litserve",559    "github_about_section": "A minimal Python framework for building custom AI inference servers with full control over logic, batching, and scaling.",560    "homepage_link": "https://lightning.ai/litserve"561  },562  {563    "repo_name": "openzl",564    "repo_link": "https://github.com/facebook/openzl",565    "category": "data compression",566    "github_about_section": "A novel data compression framework",567    "homepage_link": "https://openzl.org"568  },569  {570    "repo_name": "torchforge",571    "repo_link": "https://github.com/meta-pytorch/torchforge",572    "category": "reinforcement learning",573    "github_about_section": "PyTorch-native post-training at scale",574    "homepage_link": "https://meta-pytorch.org/torchforge"575  },576  {577    "repo_name": "torchcodec",578    "repo_link": "https://github.com/meta-pytorch/torchcodec",579    "github_about_section": "PyTorch media decoding and encoding",580    "homepage_link": "https://meta-pytorch.org/torchcodec/stable/index.html"581  },582  {583    "repo_name": "quack",584    "repo_link": "https://github.com/Dao-AILab/quack",585    "category": "kernel examples",586    "github_about_section": "A Quirky Assortment of CuTe Kernels"587  },588  {589    "repo_name": "Triton-distributed",590    "repo_link": "https://github.com/ByteDance-Seed/Triton-distributed",591    "category": "distributed computing",592    "github_about_section": "Distributed Compiler based on Triton for Parallel Systems",593    "homepage_link": "https://triton-distributed.readthedocs.io"594  },595  {596    "repo_name": "ThunderKittens",597    "repo_link": "https://github.com/HazyResearch/ThunderKittens",598    "category": "parallel computing",599    "github_about_section": "Tile primitives for speedy kernels",600    "homepage_link": "https://hazyresearch.stanford.edu/blog/2024-10-29-tk2"601  },602  {603    "repo_name": "Hugging Face Kernels",604    "repo_link": "https://github.com/huggingface/kernels",605    "category": "gpu kernels",606    "github_about_section": "Load compute kernels from the Hub"607  },608  {609    "repo_name": "GEAK-agent",610    "repo_link": "https://github.com/AMD-AGI/GEAK-agent",611    "category": "agent",612    "github_about_section": "It is an LLM-based AI agent, which can write correct and efficient gpu kernels automatically."613  },614  {615    "repo_name": "ome",616    "repo_link": "https://github.com/sgl-project/ome",617    "category": "container orchestration",618    "github_about_section": "Open Model Engine (OME) — Kubernetes operator for LLM serving, GPU scheduling, and model lifecycle management.",619    "homepage_link": "https://ome-projects.github.io/ome/"620  },621  {622    "repo_name": "OLMo-core",623    "repo_link": "https://github.com/allenai/OLMo-core",624    "category": "training framework",625    "github_about_section": "PyTorch building blocks for the OLMo ecosystem",626    "homepage_link": "https://olmo-core.readthedocs.io/en/latest/"627  },628  {629    "repo_name": "triSYCL",630    "repo_link": "https://github.com/triSYCL/triSYCL",631    "category": "parallel computing",632    "github_about_section": "Generic system-wide modern C++ for heterogeneous platforms with SYCL from Khronos Group",633    "homepage_link": "https://trisycl.github.io/triSYCL/Doxygen/triSYCL/html/index.html"634  },635  {636    "repo_name": "tritonparse",637    "repo_link": "https://github.com/meta-pytorch/tritonparse",638    "category": "performance testing",639    "github_about_section": "TritonParse: A Compiler Tracer, Visualizer, and Reproducer for Triton Kernels",640    "homepage_link": "https://meta-pytorch.org/tritonparse"641  },642  {643    "repo_name": "reference-kernels",644    "repo_link": "https://github.com/gpu-mode/reference-kernels",645    "category": "kernel examples",646    "github_about_section": "Official Problem Sets / Reference Kernels for the GPU MODE Leaderboard!",647    "homepage_link": "https://gpumode.com"648  },649  {650    "repo_name": "hatchet",651    "repo_link": "https://github.com/LLNL/hatchet",652    "category": "performance testing",653    "github_about_section": "Graph-indexed Pandas DataFrames for analyzing hierarchical performance data",654    "homepage_link": "https://llnl-hatchet.readthedocs.io"655  },656  {657    "repo_name": "kernelbot",658    "repo_link": "https://github.com/gpu-mode/kernelbot",659    "category": "kernel examples",660    "github_about_section": "Write a fast kernel and see how you compare against the best humans and AI on gpumode.com",661    "homepage_link": "https://www.gpumode.com"662  },663  {664    "repo_name": "cutile-python",665    "repo_link": "https://github.com/NVIDIA/cutile-python",666    "category": "parallel computing",667    "github_about_section": "cuTile is a programming model for writing parallel kernels for NVIDIA GPUs",668    "homepage_link": "https://docs.nvidia.com/cuda/cutile-python"669  },670  {671    "repo_name": "FTorch",672    "repo_link": "https://github.com/Cambridge-ICCS/FTorch",673    "category": "middleware",674    "github_about_section": "A library for directly calling PyTorch ML models from Fortran.",675    "homepage_link": "https://cambridge-iccs.github.io/FTorch"676  },677  {678    "repo_name": "KernelBench",679    "repo_link": "https://github.com/ScalingIntelligence/KernelBench",680    "category": "benchmark",681    "github_about_section": "KernelBench: Can LLMs Write GPU Kernels? - Benchmark with Torch -> CUDA problems",682    "homepage_link": "https://scalingintelligence.stanford.edu/blogs/kernelbench"683  },684  {685    "repo_name": "NVSHMEM",686    "repo_link": "https://github.com/NVIDIA/nvshmem",687    "category": "distributed computing",688    "github_about_section": "GPU-initiated communication for scalable NVIDIA GPU clusters.",689    "homepage_link": "https://developer.nvidia.com/nvshmem"690  },691  {692    "repo_name": "flashinfer-bench",693    "repo_link": "https://github.com/flashinfer-ai/flashinfer-bench",694    "category": "benchmark",695    "github_about_section": "Building the Virtuous Cycle for AI-driven LLM Systems",696    "homepage_link": "https://bench.flashinfer.ai"697  },698  {699    "repo_name": "Primus-Turbo",700    "repo_link": "https://github.com/AMD-AGI/Primus-Turbo",701    "category": "training framework",702    "github_about_section": "Primus-Turbo is a high-performance acceleration library dedicated to large-scale model training on AMD GPUs. Built and optimized for the AMD ROCm platform, it covers the full training stack — including core compute operators (GEMM, Attention, GroupedGEMM), communication primitives, optimizer modules, low-precision computation (FP8), and compute–communication overlap kernels."703  },704  {705    "repo_name": "BitBLAS",706    "repo_link": "https://github.com/microsoft/BitBLAS",707    "category": "Basic Linear Algebra Subprograms (BLAS)",708    "github_about_section": "BitBLAS is a library to support mixed-precision matrix multiplications, especially for quantized LLM deployment."709  },710  {711    "repo_name": "kernels-community",712    "repo_link": "https://github.com/huggingface/kernels-community",713    "category": "gpu kernels",714    "homepage_link": "https://huggingface.co/kernels-community",715    "github_about_section": "Kernel sources for https://huggingface.co/kernels-community"716  },717  {718    "repo_name": "omnitrace",719    "repo_link": "https://github.com/ROCm/omnitrace",720    "category": "performance testing",721    "github_about_section": "Omnitrace: Application Profiling, Tracing, and Analysis",722    "homepage_link": "https://rocm.docs.amd.com/projects/omnitrace"723  },724  {725    "repo_name": "PipelineRL",726    "repo_link": "https://github.com/ServiceNow/PipelineRL",727    "category": "reinforcement learning",728    "github_about_section": "A scalable asynchronous reinforcement learning implementation with in-flight weight updates.",729    "homepage_link": "https://arxiv.org/abs/2509.19128"730  },731  {732    "repo_name": "kraken",733    "repo_link": "https://github.com/meta-pytorch/kraken",734    "category": "kernel examples",735    "github_about_section": "Triton-based Symmetric Memory operators and examples"736  },737  {738    "repo_name": "TileIR",739    "repo_link": "https://github.com/microsoft/TileIR",740    "category": "parallel computing dsl",741    "github_about_section": "TileIR (tile-ir) is a concise domain-specific IR designed to streamline the development of high-performance GPU/CPU kernels (e.g., GEMM, Dequant GEMM, FlashAttention, LinearAttention). By employing a Pythonic syntax with an underlying compiler infrastructure on top of TVM, TileIR allows developers to focus on productivity without sacrificing the low-level optimizations necessary for state-of-the-art performance."742  },743  {744    "repo_name": "intelliperf",745    "repo_link": "https://github.com/AMDResearch/intelliperf",746    "category": "performance testing",747    "github_about_section": "Automated bottleneck detection and solution orchestration",748    "homepage_link": "https://arxiv.org/html/2508.20258v1"749  },750  {751    "repo_name": "tilus",752    "repo_link": "https://github.com/NVIDIA/tilus",753    "category": "parallel computing",754    "github_about_section": "Tilus is a tile-level kernel programming language with explicit control over shared memory and registers.",755    "homepage_link": "https://nvidia.github.io/tilus"756  },757  {758    "repo_name": "gemlite",759    "repo_link": "https://github.com/dropbox/gemlite",760    "category": "gpu kernels",761    "github_about_section": "Fast low-bit matmul kernels in Triton"762  },763  {764    "repo_name": "TritonBench",765    "repo_link": "https://github.com/thunlp/TritonBench",766    "category": "benchmark",767    "github_about_section": "TritonBench: Benchmarking Large Language Model Capabilities for Generating Triton Operators",768    "homepage_link": "https://arxiv.org/abs/2502.14752"769  },770  {771    "repo_name": "triton-runner",772    "repo_link": "https://github.com/toyaix/triton-runner",773    "category": "debugger",774    "github_about_section": "Multi-Level Triton Runner supporting Python, IR, PTX, and cubin.",775    "homepage_link": "https://triton-runner.org"776  },777  {778    "repo_name": "Megatron-LM",779    "repo_link": "https://github.com/NVIDIA/Megatron-LM",780    "github_about_section": "Ongoing research training transformer models at scale",781    "homepage_link": "https://docs.nvidia.com/megatron-core/developer-guide/latest/index.html"782  },783  {784    "repo_name": "fairscale",785    "repo_link": "https://github.com/facebookresearch/fairscale",786    "github_about_section": "PyTorch extensions for high performance and large scale training."787  },788  {789    "repo_name": "ColossalAI",790    "repo_link": "https://github.com/hpcaitech/ColossalAI",791    "github_about_section": "Making large AI models cheaper, faster and more accessible",792    "homepage_link": "https://colossalai.org/"793  },794  {795    "repo_name": "NeMo RL",796    "repo_link": "https://github.com/nvidia-nemo/rl",797    "github_about_section": "Scalable toolkit for efficient model reinforcement",798    "homepage_link": "https://docs.nvidia.com/nemo/rl/latest/index.html"799  },800  {801    "repo_name": "slime",802    "repo_link": "https://github.com/THUDM/slime",803    "github_about_section": "slime is an LLM post-training framework for RL Scaling.",804    "homepage_link": "https://thudm.github.io/slime/"805  },806  {807    "repo_name": "RAGEN",808    "repo_link": "https://github.com/mll-lab-nu/RAGEN",809    "github_about_section": "RAGEN leverages reinforcement learning to train LLM reasoning agents in interactive, stochastic environments.",810    "homepage_link": "https://ragen-ai.github.io/"811  },812  {813    "repo_name": "slurm",814    "repo_link": "https://github.com/SchedMD/slurm",815    "github_about_section": "Slurm: A Highly Scalable Workload Manager",816    "homepage_link": "https://slurm.schedmd.com/"817  },818  {819    "repo_name": "Open Thoughts",820    "repo_link": "https://github.com/open-thoughts/open-thoughts",821    "github_about_section": "Fully open data curation for reasoning models",822    "homepage_link": "https://www.open-thoughts.ai/"823  },824  {825    "repo_name": "Optimum",826    "repo_link": "https://github.com/huggingface/optimum",827    "github_about_section": "Accelerate inference and training of Transformers, Diffusers, TIMM and Sentence Transformers with easy to use hardware optimization tools",828    "homepage_link": "https://huggingface.co/docs/optimum/main/en/index"829  },830  {831    "repo_name": "ultralytics",832    "repo_link": "https://github.com/ultralytics/ultralytics",833    "github_about_section": "Ultralytics YOLO26, YOLO11, YOLOv8 — object detection, instance segmentation, semantic segmentation, image classification, pose estimation, object tracking",834    "homepage_link": "https://platform.ultralytics.com"835  },836  {837    "repo_name": "NVIDIA FLARE",838    "repo_link": "https://github.com/nvidia/nvflare",839    "github_about_section": "NVIDIA Federated Learning Application Runtime Environment",840    "homepage_link": "https://nvidia.github.io/NVFlare/"841  },842  {843    "repo_name": "pip",844    "repo_link": "https://github.com/pypa/pip",845    "github_about_section": "The Python package installer",846    "homepage_link": "https://pip.pypa.io/"847  },848  {849    "repo_name": "Maven",850    "repo_link": "https://github.com/apache/maven",851    "github_about_section": "Apache Maven is a software project management and comprehension tool. Based on the concept of a project object model (POM), Maven can manage a project's build, reporting and documentation from a central piece of information.",852    "homepage_link": "https://maven.apache.org"853  },854  {855    "repo_name": "SpecForge",856    "repo_link": "https://github.com/sgl-project/SpecForge",857    "github_about_section": "Train speculative decoding models effortlessly and port them smoothly to SGLang serving.",858    "homepage_link": "https://sgl-project.github.io/SpecForge/"859  }860]