harishanand95

Stable Diffusion optimized for AMD RDNA2 GPUs using IREE->Vulkan

This guide is meant for developers who want to experiment with the IREE/MLIR toolchain to compile Vulkan code. A more streamlined installation procedure (that uses precompiled .vmfb files optimized for various GPUs) would be possible, but in the interest of time, we’re sharing here instructions meant for developers.

Install the latest AMD Drivers

AMD Software: Adrenalin Edition 22.11.1 for MLIR/IREE Driver Version 22.20.29.09 for Windows® 10 and Windows® 11 (Windows Driver Store Version 31.0.12029.9003)

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-22-11-1-mril-iree

harishanand95

Stable Diffusion for AMD GPUs on Windows using DirectML

UPDATE: A faster (20x) approach for running Stable Diffusion using MLIR/Vulkan/IREE is available on Windows:

https://github.com/nod-ai/SHARK/blob/main/shark/examples/shark_inference/stable_diffusion/stable_diffusion_amd.md

Install 🤗 diffusers

conda create --name sd39 python=3.9 -y

harishanand95

cmake_minimum_required(VERSION 2.8.3 FATAL_ERROR)
cmake_policy(SET CMP0048 NEW)
cmake_policy(SET CMP0054 NEW)

if (NOT CMAKE_INSTALL_PREFIX)
  set(CMAKE_INSTALL_PREFIX "/usr" CACHE STRING "install prefix" FORCE)
endif()

message(STATUS "install-prefix: ${CMAKE_INSTALL_PREFIX}")

harishanand95

html, body {
  font-family: 'Roboto', 'Helvetica', sans-serif;
  margin: 0;
  padding: 0;
}

.mdl-demo .mdl-layout__header-row {
  padding-left: 40px;
}
.mdl-demo .mdl-layout.is-small-screen .mdl-layout__header-row h3 {

harishanand95

/* A JS version of time conversion used in systemd
* Available C version at systemd/src/basic/time-util.c
* Input t = real time microsecond (1463509800000000)
*/
function format_timestamp_relative(t) {
    var MSEC_PER_SEC = 1000;
    var USEC_PER_SEC = 1000000;
    var USEC_PER_MSEC = 1000;
    var USEC_INFINITY =-1;