Fixed- Torch is not able to use GPU AMD Stable Diffusion --skip-torch-cuda-test (AUTOMATIC1111) Bug

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hi everyone welcome to the channel currently when you build stable diffusion from the automatic d11 gab Breo with AMD support and you follow the instructions as per this video you currently end up with an error when you run the web ui- user dobat file now this error on the repo can be fixed at any day so only follow the steps in this video after following all the steps accurately in my last build video if you did follow all the steps in the previous video and you're currently running web ui- user dobat and you end up getting the following a runtime error torch is not able to use the GPU add the following to your commandline arguments skip torch Cuda test then this video might be able to help you out it seems this repo reverted to Cuda and the issue started mid December hopefully it will be fixed soon for now please please try the following steps and see if it resolves the issue open your Windows Explorer and over to the folder where you installed and configured stable diffusion we're looking for the stable diffusion config files now there's two files that we need to edit let's edit requirements uncore version.txt first edit the file and at the bottom of the file add torch D direct ml click save now edit the web ui- user dobat file add D- use- direct ml to the command line arguments this will force it to use direct ml now if you want to run natively you're good to go and you can later come back and add and optimize your command line Arguments for memory requirements and if you want to build stable diffusion with O nnx and Olive support just add D- nnx to the end of your command on line arguments save the file and let's move on to the next step open a command prompt in admin mode change directory into your stable diffusion folder we want to execute the following command then V back SL scripts back slash activate execute this command and make sure the environment activated it should be showing in brackets at the bottom left hand corner of the command prompt window and now that we confirm the enironment is activated let's quickly upgrade pup first go ahead and execute the following command python.exe space- m space pup space install space-- upgrade space pip go ahead execute this command to make sure pip upgrad successfully now with Pip upgraded we want to install httpx execute the following command pep space install space httpx equals equals 0.241 enter this and make sure httpx installed successfully and now we're ready to install the requirements execute the following command pip space install space- r space requirements.txt execute this now you should be in a position to execute the web ui- us us. bad file again this will compile and buildt stable diffusion on your PC and if all is successful it should launch your browser and land on the stable diffusion web interface and you should now have a functioning stable diffusion again either natively or with O nnx and Olive support let me know in the comments if this procedure worked for you I hope you found value in the video and hopefully it got your stable diffusion build unstuck if so give us a thumbs up a comment or consider subscribing to our Channel and keep our eye out on our channel for future Ai and Technology videos
Info
Channel: StuffAboutStuff
Views: 12,790
Rating: undefined out of 5
Keywords: Torch is not able to use GPU, --skip-torch-cuda-test, Torch is not able to use GPU; add --skip-torch-cuda-test, automatic1111 amd, automatic1111 amd bug, webui-user.bat amd, stable diffusion not working on amd cards, stable diffusion errors, torch cant use gpu amd automatic1111, automatic1111 amd not working, fix amd cuda error stable diffusion, stable diffusion cuda error, fix amd cuda issue torch, amd torch cuda problem, how to fix current broken 1111 amd repo, fix cuda amd
Id: TJ98hAIN5io
Channel Id: undefined
Length: 4min 10sec (250 seconds)
Published: Tue Jan 16 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.