#Fp8

Showing 3 of 3 repositories tagged #fp8, ranked by stars

A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.

Score

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

Microsoft Automatic Mixed Precision Library

Score

★ 636 ⑂ 50 —

Python