FICHA · MANJARO

python-triton

Language and compiler for writing highly efficient custom Deep-Learning primitives

  • Library
  • Dev
  • SCIENCE
  • HARDWARE
  • Dependency only
official+codex · reviewed · May 28, 2026 description in en

Description

Provides Triton, a language and compiler for writing custom high-performance deep-learning kernels. It helps researchers and performance engineers create optimized GPU primitives for model training and inference.

Developers use it as a specialized compute library. Kernel code depends on hardware and drivers, so validate correctness, numerical behavior, and resource use before relying on custom kernels.

Permissions

Permissions not analysed for this source yet.