AMD GPU run Fooocus on Windows! A step by step tutorial. fooocus amd

[Music] hello everyone welcome to my channel recently many people in interested in using AMD GPU on Windows to run Focus previously there was some issue in the code base so it was not able to use director ml to wrun focus using MD GBU recently I think uh thanks to the awesome contributors it is now able to use director ml to run amdgpu so in this video I want to go over the installation and also do a live run using Focus so as user my GPU is AMD 67 100 XT so as I show you right here so it has 12 GB of dedicated GPU memory V Ram so it will be used uh one thing I want to mention is that uh you do need to have a bigger memory the ram so in this case I have 20 have 32 gigabytes so usually in this machine I have like a 16 GB however it was really slow so I changed it to 32 and it was able to run Focus kind of fine so I suggest that if you have at least 32 GB of RAM that will be best to try it okay so and for CPU side do not need to have a really powerful CPU so in my case I have like a a very relatively low end CPU so that should be okay all right so so um so so for the installation side one thing you do need to you can try to follow the installation in the GitHub rapper so if you go to the AMD GPU Windows side you can try to follow this one I think that should also work for me I like to install it like a manually so I usually use a python virtual environment to do that so so in my case I usually like a get clone the repository and then I will do the for example here you would usually need to create a python V environment and specify the python version yeah and uh you will just create the environment all right so after that you can activate that so the first thing after you get a colon the repository you get a colum this reparatory yeah and and then you uh needed to go into that right so for me since I already did that I will just go to my local Repository yeah all right so once you go into the focus you can install the direct ml so what you're going to need to do is you install it first and then after that you can install the requirements all right so so after that I think you you should have everything all set and uh to start the focus you're going to need to do is to to use this command line python started and press enter so so this way it will start the program so first thing if you don't have the the model file which will need to download that so it is a bigger file it's about 6.6 GB of uh volume so so make sure that uh you have a good internet connection so so I will show you here so check points so this one will be downloaded is about 6.9 gabes this file so let's uh kick off the process press enter yeah I will also show my uh like a usage of RAM on the GPU okay so the program started the okay yeah and uh let me let me open the browser so it's like here so I think uh one thing I want to mention is that even though here it just says the total V Ram is only 1 gab that's fine so I think it's something related to the direct ml so direct ml thinks that there's only 1 GB of V Ram but uh as you will see later so all the GPU we we Ram the 12 GB will be used so so don't worry about that so I was a little worri but then I found that it doesn't matter so okay so let's give it a quick try so one thing I want to Al mention that you you can click the advanced and uh you can do on normal speed is 20 steps but uh to be honest 6700 XT it's kind a little slow like for 20 steps so I would recommend that you choose the extreme speed so this W will use the LCM ler to speed up so you will only need about five or eight steps to do that so like for my demo purose I will use the extreme speed and I I also use um some Dimension we can also use some lower Dimension I'll use this one yeah so I will use only one image all right so so I think the others can leave them as default so so all right so let's uh yeah so let's uh get it going I will have my file here and the Press generate yeah let's look at what's going on here all right we see the progress SP is uh looks good so it's need about eight steps talk about uh around the half minute yeah I don't I don't really think that's a a very fast but I think it's like uh doable right so I I don't think it's just too bad and also not really too good right so but I think it's really a good start that since the repository has enabled AMD GPU users to use it on Windows I think that's a really good start we can expect to be faster in future okay okay so at the meantime I think uh I really recommend that you use Linux so I think you can use the rock M you use the AMD software stack on Linux say that can give you some speed up so if you are interested please check out my other videos I do have some of those demos there I will add the link to the description thank you very much I hope you enjoy my video I'm very really very very excited because this I tried this this morning and I can't wait to share it with you I hope you are excited too please subscribe to my channel thank you for your support goodbye see you next time

Channel: Tech-Practice

Views: 10,879

Keywords: ai, cpu, gpu, amd, hardwares, stable diffusion, democratize, fooocus, google colab, nvidia

Id: IcfEWVvrzrA

Length: 10min 55sec (655 seconds)

Published: Thu Jan 11 2024