CUDA/HIP header-only library for low-precision (16 bit, 8 bit) and vectorized GPU kernel development
-
Updated
Jul 23, 2026 - C++
CUDA/HIP header-only library for low-precision (16 bit, 8 bit) and vectorized GPU kernel development
GRIT — Grouped Reduced-precision Interchange Type: a checkable 64-byte descriptor + contract for block-scaled, sparse, sub-byte tensors. Spec, conformance vectors, and 5 zero-dependency implementations (C, C++, Rust, Python, TypeScript). grit scan audits GGUF/safetensors files you already have.
Add a description, image, and links to the reduced-precision topic page so that developers can more easily learn about it.
To associate your repository with the reduced-precision topic, visit your repo's landing page and select "manage topics."