Info, suggestions, and congrats! #17
NeusZimmer
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi All,
Thanks for this implementation, at least , to me the faster as an AMD GPU user with only 4Gb, been able to run it at 512x512 ranging from 1.04 to 1.3x per it. for a non professional RX-5500 mobile version... very good work, impressive.(640x640x, 2.2 per it)
I used also stable-diffusion-directml and SHARK but yours is the best one i found, not by its breath of options, because its fast generation of samples. SHARK was close to this version performance, (only a couple of times even it was faster) , but with SHARK I'm on need to recompile models constantly, adding waiting time and using a lot of disk to store models...(and multiples error while trying to run them for first time...).
Only issue I found was with inpainting, I was in need to change the GPU to the standard AMD graphics card with more shared memory, due to OOM errors, and changing only the device for inpainting did the trick. (Edit: on OnnxDifffusersUI, not in your code, sorry my mistake)
-Suggestions: (sorry, all of them relevant for UI, not for you...).
-A script to update code from github? ( know some coding, but not used to github and not sure how to update the code without losing my own edits. (already figure out it, but anyway will be good to get an update script, i will check for modifications prior fetching new versions...)
-Maybe basic support for wildcards or prompt X/Y?
Keep the good work!
Beta Was this translation helpful? Give feedback.
All reactions