There have been no updates for a long time

lshqqytiger / stable-diffusion-webui-amdgpu

Stable Diffusion web UI

GNU Affero General Public License v3.0

1.68k stars 175 forks source link

There have been no updates for a long time #324

Closed SunGreen777 closed 4 months ago

SunGreen777 commented 7 months ago

Is there an existing issue for this?

[X] I have searched the existing issues and checked the recent builds/commits

What would your feature do ?

Seunghoon Lee Have you stopped providing support? Can you please tell me who to write to at AMD so they can contact you so they can allocate staff to your team?

Maybe they don't get the request and think that nobody needs it, let's create such a request, let AMD start thinking about comunitti.

If you have a different opinion, just delete this post. Thank you. : )

bebop210 commented 7 months ago

I concur.. It seems like all the focus is on Sd.next but that is inferior experience not optimised (I can not even load a model that's been "olived" all flash but completely useless.. Unless you're waiting for rocm like the rest of us.. But I'm still unsure whether to upgrade to torch 2.0 for directml so I have two conda envs but it seems like torch 1.3 can inpaint and enjoy Lora support but 2.0 won't work with either.. But at least they both start and load a model which is more than I can do with SD next vladmandic

lshqqytiger commented 7 months ago

There is no stable release on the upstream after recent update, there's no update of torch-directml after April and I'm busy now due to personal schedules until the end of December. AMD has already released ROCm for Windows and we have to wait for PyTorch to support it.

SunGreen777 commented 7 months ago

Seunghoon Lee Thanks!

ArchAngelAries commented 7 months ago

There is no stable release on the upstream after recent update, there's no update of torch-directml after April and I'm busy now due to personal schedules until the end of December. AMD has already released ROCm for Windows and we have to wait for PyTorch to support it.

Whoa, thought they were gonna take forever to release ROCm for windows, hopefully pytorch support for it doesn't take too long to implement.