Top NVIDIA GitHub Repositories & Tools (2026)

Low-Latency Accelerated Web Remote Desktop Streaming Platform for Self-Hosting, Containers, Kubernetes, or Cloud/HPC

Low-latency hardware-accelerated WebRTC video streaming and remote control for Raspberry Pi & NVIDIA Jetson

🏗️ Fine-tune, build, and deploy open-source LLMs easily!

Vulkan ray tracing glTF 2.0 renderer — RTX path tracer with PBR materials, AI denoising (DLSS, OptiX), and scene editing. NVIDIA open-source sample.

Java Inference Tornado toolkit: High performance LLM inference & serving for the JVM on GPUs

Nvidia GPU exporter for prometheus using nvidia-smi binary OR using NVML

Kubernetes operator for self-hosted LLM inference across a heterogeneous GPU fleet: NVIDIA CUDA, AMD Vulkan, and Apple Silicon Metal. Runtimes: llama.cpp, vLLM, TGI, mlx-server. Multi-GPU sharding, model caching, OpenAI-compatible endpoints. Apache-2.0, run across homelab and on-prem fleets, actively developed.

Multi-language agent runtime and library for execution scope management, lifecycle events, and middleware on tool and LLM calls.

Tensor Fusion is a state-of-the-art GPU virtualization and pooling solution designed to optimize GPU cluster utilization to its fullest potential.

Windows player that runs video, photos or GIFs through NVIDIA's DLSS 5 neural renderer, then compares the result with the original frame by frame. Local files or YouTube. Needs an RTX card and driver 610.47+. Community project, unsigned runtime. C++20, D3D12.

🕵️ NeMo Anonymizer: Detect and protect PII through context-aware replacement and rewriting

DLSS MFG on Ada — unlocks 3x/4x/6x frame generation on RTX 40-series cards and corrects the temporal midpoint. In-memory only.

DLSS 5 (Neural Rendering) em qualquer jogo, emulador ou video que apareca na tela - sem injetar nada no processo do jogo. Precisa de uma GeForce RTX.

This project is inspired by CMP40HX-Unlock. If you find it useful, please consider giving a star to both the original author and this repository.

ReShade add-on that makes DLSS 5 Neural Rendering cheaper: peripheral compression before the model and model passes every N frames with motion-vector reprojection

Pile graphique des jeux : DLSS 5, Multi Frame Generation, HDR RenoDX et ReShade - installes en un clic, verifies, retirables en un clic.

Superseded from DLSS 5 Neural Rendering v3.3.5, which handles mipped outputs itself. For earlier versions: supplies a single-mip DLSS output in D3D12 games whose output has a mip chain

Free Local Unity AI Motion Generate Tool.Based By Nvlab Kimodo.

KEDA External gRPC Scaler for GPU workloads - native NVML metrics via DaemonSet, no Prometheus required

https://wiki.seeedstudio.com Seeed Studio Wiki source code