NVlabs/tiny-cuda-nn

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

4536 stars · 580 forks observed · 2026-09-23

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

A small, self-contained C++/CUDA framework for training and querying neural networks, featuring a lightning-fast fully fused MLP and a versatile multiresolution hash encoding. It also provides PyTorch bindings and supports various input encodings, losses, and optimizers.

Use cases

• train tiny MLPs at maximum GPU speed
• implement NeRF-style neural rendering
• use multiresolution hash encoding for fast signal fitting
• accelerate neural network inference in real-time graphics
• integrate fast neural networks into PyTorch pipelines

When to choose

• you need extremely fast training or inference of small MLPs on NVIDIA GPUs
• you are building NeRF or instant-NGP-style neural rendering systems
• you want a lightweight C++/CUDA neural network library with PyTorch bindings

When to avoid

• you need large-scale deep learning models like transformers
• you lack an NVIDIA CUDA-capable GPU
• you need a general-purpose framework with broad ecosystem support like PyTorch or TensorFlow

Facets

library · maturity active

deep-learning machine-learning graphics gpu-computing deep-learning machine-learning graphics computer-vision cpp python windows cuda fully-fused-mlp hash-encoding nerf neural-networks pytorch-bindings real-time-training gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NVlabs/tiny-cuda-nn")