#143143rd of 168 in ML frameworks and librariesFor image, video and design, 3d
tiny-cuda-nn
CUDA library of fused small MLPs and hash encodings, with PyTorch bindings
- Stars
- 4.5k
- License
- custom license
- Last commit
- Sep 2026
- Last release
- Jul 2025
- Language
- C++
4,441 on 21 Apr 20264,547 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
tiny-cuda-nn is a C++/CUDA framework for training and querying small neural networks, built around a fully fused MLP and multiresolution hash encoding. It also ships other encodings, losses and optimizers, an optional JIT fusion mode (v2.0+) using CUDA runtime compilation, and a PyTorch extension. It underlies Instant NGP and related neural graphics work.
Who it is for: Graphics and ML engineers building small, fast neural fields on NVIDIA GPUs
Strengths
- Fully fused MLP kernels, benchmarked in the README against TensorFlow 2.5 with XLA on an RTX 3090
- JIT fusion gives a stated 1.5x to 2.5x speedup, more on newer GPUs
- Hash grid, frequency, oneblob and spherical harmonics encodings, plus Adam and Shampoo optimizers
- Usable from C++ or from PyTorch via the tinycudann package
Weaknesses
- Requires an NVIDIA GPU and CUDA toolchain; no other backends are listed
- Default fully fused MLP needs large shared memory; RTX 2080 Ti or better
- PyTorch binding is about 2x slower than native CUDA at 64k batch size
- JIT fusion can slow training on RTX 3000 or older, or with very large grids
What it needs
- GPU required
- Needs CUDA, CMake 3.21+, C++17 compiler, PyTorch (optional bindings)
- Models: FullyFusedMLP, CutlassMLP, multiresolution hash grid encoding
Quick answers
Is tiny-cuda-nn open source?
Its code is public on GitHub under its own license terms; read them before you use it at work.
Does tiny-cuda-nn need a GPU?
Yes, it needs a GPU.
Also in ML frameworks and libraries
See all 168| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd | 93 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts | 93 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 92 out of 100 |