Project Icon

triton

Enhancing Custom Deep-Learning Primitives with Triton's Flexible Compiler

Product DescriptionTriton is an open-source language and compiler aimed at creating efficient custom deep-learning primitives, offering flexibility and productivity beyond CUDA. It supports fast computational code development with comprehensive tools and documentation. Triton facilitates advanced neural computations on GPUs and CPUs, and supports NVIDIA and AMD GPUs. Easily integrate Triton using pip with LLVM customization options. Join the Triton Conference on September 17, 2024, in Fremont, CA, for community insights and developments.
Project Details