You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Sep 7, 2022. It is now read-only.
So, my Docker knowledge is even more limited than my linux knowledge but it's getting there bit by bit. So anyways...
I used https://github.com/AshleyYakeley/stable-diffusion-rocm to get a docker image running SD on my AMD 6800 gpu on CLI (On Manjaro). No tweaks needed, it just worked. (I had issues with the OG https://www.youtube.com/watch?v=d_CgaHyA_n4 that I mostly don't even remember. I installed ubuntu onto a partition and gave the vid a go. Didn't work out, then I found Ashley's dockerfile, tried it on my Manjaro install instead and it just worked.)
Woo.
I then figured out making a commit of the docker file in a new name after a initial run of SD is completed so that first time download is out of the way on following fresh runs of it.
Neat
I then combined that modified run script with the idea to add in an input folder volume by adding "-v pwd/input:/input " to the end of her script and sudo copying in a image file to the new input folder (because for some reason that is root privileged!?) so I can run img2img too.
Great.
Now I have my sights on transitioning to using this lovely looking Web UI here...
My first noob impression is that that linux-sd.sh script by and large does all the installation all over again rawdog onto my host machine?
I am not against this, my concern is that I guess it doesn't factor in replacing the CUDA stuff with ROCM and correct pytorch version etc that I need to run it on my AMD hardware.
If I want this GUI and the face and upscaling addons, do I need to use this linux-sd.sh script and figure out how to modify it to the AMD friendly version, or can I install this to the docker container that is proven to work and then just commit a new docker image with this on top of it... does that make sense?
I know if I went away and spent a good while learning more about linux, docker and all sorts of things I'd be able to formulate a better question... But all in all, for a not total beginner, does anyone have any advice for me in this situation?
Also I have read through #93 so I know this can all be "dockerized" but oc013's guide seems to be for a more vanilla Nvidia install, so is only of partial help?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
So, my Docker knowledge is even more limited than my linux knowledge but it's getting there bit by bit. So anyways...
I used https://github.com/AshleyYakeley/stable-diffusion-rocm to get a docker image running SD on my AMD 6800 gpu on CLI (On Manjaro). No tweaks needed, it just worked. (I had issues with the OG https://www.youtube.com/watch?v=d_CgaHyA_n4 that I mostly don't even remember. I installed ubuntu onto a partition and gave the vid a go. Didn't work out, then I found Ashley's dockerfile, tried it on my Manjaro install instead and it just worked.)
Woo.
I then figured out making a commit of the docker file in a new name after a initial run of SD is completed so that first time download is out of the way on following fresh runs of it.
Neat
I then combined that modified run script with the idea to add in an input folder volume by adding "-v
pwd
/input:/input " to the end of her script and sudo copying in a image file to the new input folder (because for some reason that is root privileged!?) so I can run img2img too.Great.
Now I have my sights on transitioning to using this lovely looking Web UI here...
I've taken a gander at https://github.com/hlky/stable-diffusion/wiki/Linux-Installation
My first noob impression is that that linux-sd.sh script by and large does all the installation all over again rawdog onto my host machine?
I am not against this, my concern is that I guess it doesn't factor in replacing the CUDA stuff with ROCM and correct pytorch version etc that I need to run it on my AMD hardware.
If I want this GUI and the face and upscaling addons, do I need to use this linux-sd.sh script and figure out how to modify it to the AMD friendly version, or can I install this to the docker container that is proven to work and then just commit a new docker image with this on top of it... does that make sense?
I know if I went away and spent a good while learning more about linux, docker and all sorts of things I'd be able to formulate a better question... But all in all, for a not total beginner, does anyone have any advice for me in this situation?
Also I have read through #93 so I know this can all be "dockerized" but oc013's guide seems to be for a more vanilla Nvidia install, so is only of partial help?
Beta Was this translation helpful? Give feedback.
All reactions