[Issue]: "Bad sampler name: None" while using hires fix option.

[mart-hill]

Issue Description

The error shows up like this:

02:49:24-050625 ERROR    gradio call: AssertionError
╭─────────────────────────────────────── Traceback (most recent call last) ────────────────────────────────────────╮
│ X:\AI\automatic\modules\call_queue.py:34 in f                                                                    │
│                                                                                                                  │
│    33 │   │   │   try:                                                                                           │
│ ❱  34 │   │   │   │   res = func(*args, **kwargs)                                                                │
│    35 │   │   │   │   progress.record_results(id_task, res)                                                      │
│                                                                                                                  │
│ X:\AI\automatic\modules\txt2img.py:58 in txt2img                                                                 │
│                                                                                                                  │
│   57 │   if processed is None:                                                                                   │
│ ❱ 58 │   │   processed = process_images(p)                                                                       │
│   59 │   p.close()                                                                                               │
│                                                                                                                  │
│                                             ... 4 frames hidden ...                                              │
│                                                                                                                  │
│ X:\AI\automatic\extensions\sd-webui-neutral-prompt\lib_neutral_prompt\cfg_denoiser_hijack.py:245 in              │
│ create_sampler_hijack                                                                                            │
│                                                                                                                  │
│   244 def create_sampler_hijack(name: str, model, original_function):                                            │
│ ❱ 245 │   sampler = original_function(name, model)                                                               │
│   246 │   if name.startswith(('DDIM', 'PLMS', 'UniPC')):                                                         │
│                                                                                                                  │
│ X:\AI\automatic\modules\sd_samplers.py:30 in create_sampler                                                      │
│                                                                                                                  │
│   29 │   config = find_sampler_config(name)                                                                      │
│ ❱ 30 │   assert config is not None, f'bad sampler name: {name}'                                                  │
│   31 │   if backend == Backend.ORIGINAL:                                                                         │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
AssertionError: bad sampler name: None

I wonder if the newest update fixes this - gotta test! :)

Version Platform Description

01:42:03-296111 INFO Running extension preloading 01:42:03-309111 INFO Starting SD.Next 01:42:03-311111 INFO Python 3.10.9 on Windows 01:42:03-494111 INFO Version: 562947c9 Fri Jun 2 01:14:12 2023 +0300 01:42:04-078111 INFO Setting environment tuning 01:42:04-096111 INFO nVidia CUDA toolkit detected 01:42:06-429611 INFO Torch 2.0.1+cu118 01:42:06-459114 INFO Torch backend: nVidia CUDA 11.8 cuDNN 8700 01:42:06-461616 INFO Torch detected GPU: NVIDIA GeForce RTX 3090 VRAM 24575 Arch (8, 6) Cores 82 01:42:06-463615 INFO Verifying requirements 01:42:06-487113 INFO Installing packages 01:42:06-698611 INFO Running setup 01:42:06-699611 INFO Installing repositories 01:42:13-444840 INFO Installing submodules 01:42:17-746931 INFO Updating submodules

Edit: Oh, so I'm supposed to select some sampler here? I had None here all this time.

[myndxero]

TEMPORARY BAND-AID UNTIL VLAD SEES THIS in the two backup sampler boxes, just put a space and click out so the box is 'empty,' not even having 'None' selected, and apply settings. No more issue.

This is broken for me too after just updating moments ago.

Oh, so I'm supposed to select some sampler here? I had None here all this time.

I've had this set to None forever. Same with the option below it regarding latent upscaler. Is this the culprit all of a sudden?

[mart-hill]

For me, the only way is to set the "Secondary sampler" to "Euler a" to make this error go away, for now. 🙂

"UNTIL VLAD SEES THIS" - he's gonna cry because it's me, again. :P

[vladmandic]

the app code does checks for valid samplers - if i've missed any, i'll fix it
but issue here is in extension sd-webui-neutral-prompt - it HIJAACKS sampler creation and performs NO CHECKS

i've removed assert causing runtime error and replace it with error message and fallback to default sampler
other than that, not much i can do

[mart-hill]

Thank you! Now I'll have to report the very same bug, albeit differently "asserted" (AttributeError: module 'modules.sd_samplers' has no attribute 'create_sampler_original_md'), for Tiled Diffusion extension. This function will work, if I set "Euler a" as fallback sampler. 🙂

[vladmandic]

the problem is that both extensions are trying to hijaack same function. so one hijaacks, then second hijaacks, then first one complains "oh, i'm missing something". it works if you set fallback sampler explicity since then extension skips part of its code trying to set it itself, but it doesn't mean they are not misbehaving.