ThunderKittens
GPU kernel framework
An open-source framework for efficiently writing deep learning kernels on NVIDIA GPUs.
Tile primitives for speedy kernels
2k stars
30 watching
79 forks
Language: Cuda
last commit: about 2 months ago
Linked from 1 awesome list
Related projects:
Repository | Description | Stars |
---|---|---|
xtra-computing/thundergbm | Accelerates machine learning algorithms on GPUs to improve performance and efficiency | 695 |
komputeproject/kompute | A flexible GPU compute framework providing low-level access to Vulkan for optimized and parallel processing on various graphics cards. | 2,036 |
hannes-brt/hebel | A Python library for building and training neural networks using GPU acceleration | 1,169 |
cg-tuwien/auto-vk-toolkit | A C++ framework for creating Vulkan-based graphics applications with built-in support for various features and tools. | 412 |
fsole/brokkr | A Vulkan framework for building Windows-based graphics applications using C++. | 88 |
can-lehmann/owlkettle | A declarative user interface framework built on top of GTK 4. | 385 |
glavnokoman/vuh | A Vulkan-based framework for accelerating computations on graphics processing units. | 347 |
coolbutuseless/devoutpdf | A custom PDF graphics device for R, providing fine-grained control over output and serving as a learning tool for graphics device implementation. | 8 |
coreylowman/dfdx | A deep learning library for Rust with GPU acceleration and ergonomic API. | 1,754 |
jgbit/vuda | Provides a Vulkan-based interface to CUDA's runtime API for GPU-accelerated applications | 869 |
kwotsin/tensorflow-xception | An implementation of a deep learning model for computer vision tasks using TensorFlow | 208 |
cg-tuwien/vulkanlaunchpad | A Vulkan-based framework for beginners to learn and develop 3D graphics applications. | 64 |
denizyuret/knet.jl | A deep learning framework implemented in Julia for automatic differentiation and GPU operation. | 1,431 |
michaldrobot/shaderfastlibs | Optimized shader libraries for fast operations on graphics processing units. | 359 |
keras-team/keras | A high-level deep learning framework for building and training neural networks on multiple backend engines | 62,196 |