6 ms·

It's unfortunate that this article doesn't specify the amount of VRAM needed, other than specifying it's "less than 10Gb". I have 6,1Gb of VRAM and I tried to f

It's unfortunate that this article doesn't specify the amount of VRAM needed, other than specifying it's "less than 10Gb". I have 6,1Gb of VRAM and I tried to follow the article until eventually encountering an "unable to allocate memory" error. (I'm now trying to run basujindal's repo as an alternative.)

capableweb 4y ago
Reduce the resolution and run with half-precision instead of full-precision and you should be able to avoid OOM errors. Author seems to have had 8GB VRAM available, so I'm guessing that's the "minimum required" for their solution.
- baobabKoodaa 4y ago
  It's not possible to halve the precision further. The precision was already dropped from float32 to float16 in the OP. I now used parameters to drop the resolution to 256x256, and now it's running, but it's somehow broken. Every output image it produces is literally a green square.
  - LanternLight83 4y ago
    The green square issue has been well known, particularly on AMD cards, and I believe the solution is... full precision :c But idk, I haven't had that issue. My issue's that I can run it in <4GB VRAM, but can only do a couple dozen images before some memory leak or smth drives it out of memory (effects my 2070S too, but only after many more images). Restarting it isn't too bad, but it's enough to have me looking to using either if two AMD APU's that I have on hand.
  - XorNot 4y ago
    You need to be in full precision mode in that case. Running on my AMD card this was necessary.
    - baobabKoodaa 4y ago
      Runs out of memory in full precision mode.
      - XorNot 4y ago
        Try the basujindal fork https://github.com/basujindal/stable-diffusion https://github.com/basujindal/stable-diffusion - this is what works for me on 8gb of VRAM (without maybe 6.5 available from whatever my DE is using up). On Linux with AMD `radeontop` is quite informative.
  - neurostimulant 4y ago
    I also got green square when running with half precision on gtx 1650. Try 384x384 with full precision instead, or use the intel cpu fork.
hedora 4y ago
See my other comment in this article. I'm running with 8, but it never uses more than 6GB.

capableweb 4y ago

Reduce the resolution and run with half-precision instead of full-precision and you should be able to avoid OOM errors. Author seems to have had 8GB VRAM available, so I'm guessing that's the "minimum required" for their solution.

baobabKoodaa 4y ago
It's not possible to halve the precision further. The precision was already dropped from float32 to float16 in the OP. I now used parameters to drop the resolution to 256x256, and now it's running, but it's somehow broken. Every output image it produces is literally a green square.
- LanternLight83 4y ago
  The green square issue has been well known, particularly on AMD cards, and I believe the solution is... full precision :c But idk, I haven't had that issue. My issue's that I can run it in <4GB VRAM, but can only do a couple dozen images before some memory leak or smth drives it out of memory (effects my 2070S too, but only after many more images). Restarting it isn't too bad, but it's enough to have me looking to using either if two AMD APU's that I have on hand.
- XorNot 4y ago
  You need to be in full precision mode in that case. Running on my AMD card this was necessary.
  - baobabKoodaa 4y ago
    Runs out of memory in full precision mode.
    - XorNot 4y ago
      Try the basujindal fork https://github.com/basujindal/stable-diffusion https://github.com/basujindal/stable-diffusion - this is what works for me on 8gb of VRAM (without maybe 6.5 available from whatever my DE is using up). On Linux with AMD `radeontop` is quite informative.
- neurostimulant 4y ago
  I also got green square when running with half precision on gtx 1650. Try 384x384 with full precision instead, or use the intel cpu fork.

hedora 4y ago

See my other comment in this article. I'm running with 8, but it never uses more than 6GB.