Closed cgidesign-de closed 5 months ago
with adetailer it actually sets masks and triggers processing in the main app with newly set params. but for batch-count>1, it looks like it doesn't set mask since it thinks processing is over, so what happens is that run on second batch is without mask being set thus causing error.
i have to admit, i'm getting a bit lost in adetailer internal logic, but problem seems to be around this function: https://github.com/Bing-su/adetailer/blob/3f1d1b9772aae767ff63be15be425348b0083324/scripts/!adetailer.py#L758
this should be reported upstream to adetailer. but given the past comments, not sure if batching is considered supported at all by adetailer - see https://github.com/Bing-su/adetailer/issues/434#issuecomment-1845232897
Thanks, I posted in there as well.
Bing-Su answered in the issue thread. He wrote, Batch-Mode (multiple images in one run) is not supported, but Batch-Count is ok. As of his comment, it works in webUI 1.8.0 with image2image.
that may be so, but i cannot trace it down within adetailer code to see if i a fix is needed on sdnext side, this is extremely convoluted from adetailer side.
Ok, pity, but that's life.
i've added a workaround for this issue. its not ideal, but its something since adetailer author is not going to look at it.
I tried with --upgrade flag to get latest sd.next - issue is still there.
We have to wait for a new release, right?
right. or use dev branch.
I have upgraded today - now it is working. Thanks for solving this :-)
Issue Description
Steps to reproduce:
Image creation starts and ADetailer starts finding the face and is creating the new face details. Once this process is finished the error occures - see attached log. If Batchcount is set to 1 instead of 2 the error does not occure and the final image gets created successfully.
In Text2Image both cases work without error - Batchcount =1 or Batchcount = 2 are both finishing successfully.
LOG:
Version Platform Description
app: SD.next updated: 2024-02-24 hash: c1dfb1b2 url: https://github.com/vladmandic/automatic/tree/master
arch: AMD64 cpu: Intel64 Family 6 Model 183 Stepping 1, GenuineIntel system: Windows release: Windows-10-10.0.22631-SP0 python: 3.10.11
Torch: 2.2.0+cu121 Autocast half
GPU: device: NVIDIA GeForce RTX 4070 Ti SUPER (1) (sm_90) (8, 9) cuda: 12.1 cudnn: 8801 driver: 551.23
xformers: diffusers: 0.26.3 transformers: 4.37.2
device: active: cuda dtype: torch.float16 vae: torch.float16 unet: torch.float16
ADetailer: Created Wed Apr26 2023 07:54 | Added Fri May12 2023 00:00 | Pushed Tue Feb27 2024 12:58 | Updated Tue Jan23 2024 13:20
Browser: Firefox 123
URL link of the extension
https://github.com/Bing-su/adetailer
URL link of the issue reported in the extension repository
No response
Acknowledgements