ferrum-kernels 0.7.7

Unified compute kernels (CUDA/Metal/CPU) and model runner for Ferrum inference
Documentation 
1
2
3
4

// This is purely so that it works with torch 2.1. For torch 2.2+ we can include ATen/cuda/PhiloxUtils.cuh

#pragma once
#include <ATen/cuda/detail/UnpackRaw.cuh>