[Bug]: SD upscale error on second tile

Question

[Bug]: SD upscale error on second tile

Ladypoly opened this issue 2 years ago · 17 comments

Ladypoly commented 2 years ago

Is there an existing issue for this?

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

What happened?

SD upscale stops after finishing the first tile and throws an error.

Steps to reproduce the problem

Go to img2img
Select SD Upscale
Select SwinIR_4x

What should have happened?

Generate all necessary tiles.

Commit where the problem happens

d35bf64

What platforms do you use to access UI ?

Windows

What browsers do you use to access the UI ?

Mozilla Firefox

Command Line Arguments

"--ckpt-dir=D:\StableDiffusion/SD_Models" --xformers --deepdanbooru "--vae-path=D:\StableDiffusion/SD_Models/ft-MSE/vae-ft-mse-840000-ema-pruned.vae.pt" --api --autolaunch

Additional information, context and logs

SD upscale throws a error after completing the first tile.

SwinIR tiles: 100%|████████████████████████████████████████████████████████████████████| 15/15 [00:12<00:00, 1.19it/s]
SD upscaling will process a total of 9 images tiled as 3x3 per upscale in a total of 9 batches.
100%|██████████████████████████████████████████████████████████████████████████████████| 23/23 [00:03<00:00, 7.07it/s]
Error completing request████▏ | 23/207 [00:03<00:25, 7.26it/s]
Arguments: (0, 'A (cyberpunk :1.2)castle with a river in front and fireworks in the sky, 8 k, unreal engine, octane, trending on artstation', '', 'None', 'None', <PIL.Image.Image image mode=RGB size=832x512 at 0x1A560473DF0>, None, None, None, 0, 150, 0, 4, 0, False, False, 1, 1, 7, 0.15, 4277678666.0, -1.0, 0, 0, 0, False, 512, 832, 0, False, 32, 0, '', '', 13, 0.9, 5, '0.0001', False, 'None', '', 0.1, False, 'Not set', 'Not set', 'Not set', 'Not set', 'No focus', '', True, False, '

CFG Scale should be 2 or lower.

\n', True, True, '', '', True, 50, True, 1, 0, False, None, '', 0.2, 0.1, 1, 1, False, True, True, False, False, False, False, 4, 1, 4, 0.09, True, 1, 0, 7, False, False, 0, 0, 'Show/Hide Canvas', 0, '

Recommended settings: Sampling Steps: 80-100, Sampler: Euler a, Denoising strength: 0.8

', 128, 8, ['left', 'right', 'up', 'down'], 1, 0.05, 128, 4, 0, ['left', 'right', 'up', 'down'], False, False, None, '', '

Will upscale the image to twice the dimensions; use width and height sliders to set tile size

', 64, 8, 10.0, True, 30.0, True, 'Illustration', 'svg', True, True, False, 0.5, 1, '', 0, '', True, True, False, 1.0, 2.0, 'a painting in', 'style', 'picture frame, portrait photo', None) {}

Traceback (most recent call last):
File "D:\StableDiffusion\stable-diffusion-webui\modules\ui.py", line 185, in f
res = list(func(*args, **kwargs))
File "D:\StableDiffusion\stable-diffusion-webui\webui.py", line 54, in f
res = func(*args, **kwargs)
File "D:\StableDiffusion\stable-diffusion-webui\modules\img2img.py", line 135, in img2img
processed = modules.scripts.scripts_img2img.run(p, *args)
File "D:\StableDiffusion\stable-diffusion-webui\modules\scripts.py", line 282, in run
processed = script.run(p, *script_args)
File "D:\StableDiffusion\stable-diffusion-webui\scripts\sd_upscale.py", line 75, in run
processed = processing.process_images(p)
File "D:\StableDiffusion\stable-diffusion-webui\modules\processing.py", line 427, in process_images
res = process_images_inner(p)
File "D:\StableDiffusion\stable-diffusion-webui\modules\processing.py", line 487, in process_images_inner
with torch.no_grad(), p.sd_model.ema_scope():
AttributeError: 'NoneType' object has no attribute 'ema_scope'

Answer 1 · 2022-11-01T12:11:27.000Z

same problem with img2tiles
AttributeError: 'NoneType' object has no attribute 'ema_scope'`
error after update

Answer 2 · 2022-11-01T12:19:16.000Z

img2img loopback has the same problem

Answer 3 · 2022-11-01T13:57:27.000Z

Batch img2img has the same problem

Answer 4 · 2022-11-01T14:03:22.000Z

I can't test right now but I'm going to go out on a limb that f126986 is a probable reason for this. The error is happening in processing.py and this is the only change to processing.py in the past day.

File "D:\StableDiffusion\stable-diffusion-webui\modules\processing.py", line 487, in process_images_inner
with torch.no_grad(), p.sd_model.ema_scope():
AttributeError: 'NoneType' object has no attribute 'ema_scope'

Changes made to processing.py with that commit.

     if p.scripts is not None:
         p.scripts.postprocess(p, res)

+    p.sd_model = None
+    p.sampler = None

    return res

Can't test myself, but try going to a commit (git checkout 5b0f624) before this one. Alternatively commenting out those 2 lines in processing.py might resolve those features not working.

Quoting the PR:

PS: I confirmed that the x/y plot still adds and keeps references, but I couldn't find what part of the code was keeping the references. I cannot run all features, but the basic workflow of txt2img/img2img > change model to inpaint > inpaint > change model back is working without leak.

Answer 5 · 2022-11-01T14:12:27.000Z

Not sure of repercussions elsewhere, but commenting out these two lines gets batch img2img working again:

p.sd_model = None
p.sampler = None

Answer 6 · 2022-11-01T17:15:57.000Z

I'm getting this issue with the Wildcard and inpainting scripts as well.

Answer 7 · 2022-11-01T17:34:19.000Z

getting this error when running loopback

AttributeError: 'NoneType' object has no attribute 'ema_scope'

Commit hash: 198a1ff

Answer 8 · 2022-11-01T19:52:20.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

Answer 9 · 2022-11-01T20:04:07.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

Sharing the error output you get might help, there may be more issues beyond just those two lines.

Answer 10 · 2022-11-01T20:14:31.000Z

Loopback Error:
Traceback (most recent call last):
File "C:\Users\AAAA\stable-diffusion-webui\modules\ui.py", line 185, in f
res = list(func(*args, **kwargs))
File "C:\Users\AAAA\stable-diffusion-webui\webui.py", line 54, in f
res = func(*args, **kwargs)
File "C:\Users\AAAA\stable-diffusion-webui\modules\img2img.py", line 136, in img2img
processed = modules.scripts.scripts_img2img.run(p, *args)
File "C:\Users\AAAA\stable-diffusion-webui\modules\scripts.py", line 282, in run
processed = script.run(p, *script_args)
File "C:\Users\AAAA\stable-diffusion-webui\scripts\loopback.py", line 62, in run
processed = processing.process_images(p)
File "C:\Users\AAAA\stable-diffusion-webui\modules\processing.py", line 427, in process_images
res = process_images_inner(p)
File "C:\Users\AAAA\stable-diffusion-webui\modules\processing.py", line 487, in process_images_inner
with torch.no_grad(), p.sd_model.ema_scope():
AttributeError: 'NoneType' object has no attribute 'ema_scope'

Answer 11 · 2022-11-01T20:23:36.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

yes, just # comment out the two lines in \modules\processing.py

p.sd_model = None

p.sampler = None

loopback now working

Answer 12 · 2022-11-01T20:44:09.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

Sharing the error output you get might help, there may be more issues beyond just those two lines.

I'm getting the same messages as what Medelisa posted, and pretty close to OP's (different prompt and stuff).

Answer 13 · 2022-11-01T20:45:59.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

Sharing the error output you get might help, there may be more issues beyond just those two lines.

I'm getting the same messages as what Medelisa posted, and pretty close to OP's (different prompt and stuff).

make sure you are not updating the repo (disable autp update / git pull ) everytime you change the processing,py and restarting webUI. ( u need to restart the program )

Answer 14 · 2022-11-01T20:55:18.000Z

Had saved and restarted countless times before; somehow this time it seems to have worked....

Answer 15 · 2022-11-02T01:11:16.000Z

I made a mistake in a previous PR, I just sent a new PR #4142 to solve it, sorry for the trouble.

Answer 16 · 2022-11-03T22:09:11.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

yes, just # comment out the two lines in \modules\processing.py

p.sd_model = None

p.sampler = None

loopback now working

So I have that problem in batch and i tried doing this. I tried the # just before the p, one space away and at the beginning of the line because I'm not a programmer.
However, I end up with another error when trying to run batch img2img:
AttributeError: 'NoneType' object has no attribute 'getexif'

Answer 17 · 2022-11-24T00:32:39.000Z

Has anyone got loopback to work? I tried commenting those 2 lines but doesn't seem to be having any effect. Not sure if I'm doing something wrong...

yes, just # comment out the two lines in \modules\processing.py

p.sd_model = None

p.sampler = None

loopback now working

I am a complete newbie. How can I edit these files? I can not find this modules/processing.py file anywhere.