progress bar in SD.Next

[valery016]

I tried SD.Next webui a couple of times under Stability Matrix, it is a very appealing SD platform. One thing about it however puzzled me each time: there appears no progress bar in its interface - the one that would show the entire progress of the generation process, not just that of the current stage (e.g. VAE encoding). Maybe I am missing something?

[vladmandic]

and how would you represent that? for example, vae decode may take a long time, but its a single operation and there is no way to see inside of it how much of decode is complete or not.

[valery016]

Well, you could certainly have arranged a basic progress-bar tracking for tile-based processes, like SD Upscale / Ultimate SD Upscale (those that take the longest), couldn’t you? What I see in SD.Next’s UI, it just sits there showing ‘img2img 100% Finishing’ most of the time, until the very end. (Which takes btw a little over 5 minutes in SD.Next to 1 X SD-upscale a 7K image, as compared to just 2.5 minutes in Forge for the same image, checkpoint and all other parameters.) The console log, in the meantime, displays merrily a tile after tile being processed, in a very verbose manner.

For a non-tiled generation process, the sampling stage takes the longest, and it’s not tracked in SD.Next’s progress bar until 40% of sampling is done (which can take quite a long time, with some samplers and parameters), by what I see looking in the console logs. I tried to prepare screenshots to illustrate that with an example of batch-processing, but suddenly SD.Next developed a problem, starting to hang right after completion of the vae decode stage (indicating ‘vae 100% finishing’ in the UI and nothing further) at the end of the batch-generating of 8 1Kx1K images. Restarting SD.Next does not help, it runs into the same problem every and each time. Here is the console log of the above operation, from rebooting till the moment it hangs, in case you want to look into the issue (GPU: nvidia RTX 4070 Ti Super, 16 GB VRAM):

13:45:48-401228 INFO Starting SD.Next
13:45:48-403236 INFO Logger: file="C:\AI\StabilityMatrix\Packages\SD.Next
Web UI\sdnext.log" level=INFO size=703003 mode=append
13:45:48-404823 INFO Python 3.10.11 on Windows
13:45:48-497988 INFO Version: app=sd.next updated=2024-05-07 hash=e081f232
branch=master
url=https://github.com/vladmandic/automatic/tree/maste
r
13:45:48-815386 INFO Latest published version:
e2af454
2024-05-21T18:51:30Z
13:45:48-828646 INFO Platform: arch=AMD64 cpu=Intel64 Family 6 Model 183
Stepping 1, GenuineIntel system=Windows
release=Windows-10-10.0.22631-SP0 python=3.10.11
13:45:48-830648 INFO nVidia CUDA toolkit detected: nvidia-smi present
13:45:48-895109 INFO Extensions: disabled=['sd-webui-controlnet']
13:45:48-895109 INFO Extensions: enabled=['Lora', 'sd-extension-chainner',
'sd-extension-system-info',
'sd-webui-agent-scheduler',
'stable-diffusion-webui-images-browser',
'stable-diffusion-webui-rembg'] extensions-builtin
13:45:48-895109 INFO Extensions:
enabled=['multidiffusion-upscaler-for-automatic1111',
'sd-webui-kohya-hiresfix',
'ultimate-upscale-for-automatic1111'] extensions
13:45:48-895109 INFO Startup: quick launch
13:45:48-895109 INFO Verifying requirements
13:45:48-895109 INFO Verifying packages
13:45:48-910738 INFO Extensions: disabled=['sd-webui-controlnet']
13:45:48-910738 INFO Extensions: enabled=['Lora', 'sd-extension-chainner',
'sd-extension-system-info',
'sd-webui-agent-scheduler',
'stable-diffusion-webui-images-browser',
'stable-diffusion-webui-rembg'] extensions-builtin
13:45:48-910738 INFO Extensions:
enabled=['multidiffusion-upscaler-for-automatic1111',
'sd-webui-kohya-hiresfix',
'ultimate-upscale-for-automatic1111'] extensions
13:45:48-910738 INFO Command line args: ['--use-cuda'] use_cuda=True
13:45:52-306851 INFO Load packages: {'torch': '2.2.0+cu121', 'diffusers':
'0.27.0', 'gradio': '3.43.2'}
13:45:53-057816 INFO Engine: backend=Backend.DIFFUSERS compute=cuda
device=cuda attention="Disabled" mode=no_grad
13:45:53-107076 INFO Device: device=NVIDIA GeForce RTX 4070 Ti SUPER n=1
arch=sm_90 cap=(8, 9) cuda=12.1 cudnn=8801
driver=552.22
13:45:53-682843 INFO Available VAEs: path="models\VAE" items=0
13:45:53-683846 INFO Disabled extensions: ['sd-webui-controlnet',
'multidiffusion-upscaler-for-automatic1111']
13:45:53-684849 INFO Removing empty folder: models\BSRGAN
13:45:53-690845 INFO Available models: path="models\Stable-diffusion"
items=45 time=0.00
13:45:53-752397 INFO Extension:
script='extensions-builtin\Lora\scripts\lora_script.py
' 13:45:53-752397 INFO
LoRA networks: available=60
folders=2
13:45:54-077658 INFO Extension:
script='extensions-builtin\sd-webui-agent-scheduler\sc
ripts\task_scheduler.py' Using sqlite file:
extensions-builtin\sd-webui-agent-scheduler\task_sched
uler.sqlite3
13:45:54-249146 ERROR Upscaler unrecognized model: type=RealESRGAN
model=RealESRGAN realesrganX4plusAnime_v1
13:45:54-249146 ERROR Upscaler unrecognized model: type=RealESRGAN
model=RealESRGAN realesrganX4plus_v1
13:45:54-264149 ERROR Upscaler unrecognized model: type=RealESRGAN
model=RealESRGAN realesrGeneralX4_v3
13:45:54-264149 INFO UI theme: name="light-teal" style=Auto base=sdnext.css
13:45:55-141403 INFO Extension list is empty: refresh required
13:45:55-925254 INFO Local URL: http://127.0.0.1:7861/
13:45:56-185636 INFO [AgentScheduler] Task queue is empty
13:45:56-186623 INFO [AgentScheduler] Registering APIs
13:45:56-257523 INFO Select: model="juggernautXL_v9Rundiffusionphoto2
[c9e3e68f89]"
13:45:56-289541 INFO Setting Torch parameters: device=cuda
dtype=torch.float16 vae=torch.float16
unet=torch.float16 context=no_grad fp16=True bf16=None
optimization=Disabled
13:45:56-289541 INFO Autodetect: model="Stable Diffusion XL"
class=StableDiffusionXLPipeline
file="C:\AI\StabilityMatrix\Packages\SD.Next Web
UI\models\Stable-diffusion\juggernautXL_v9Rundiffusion
photo2.safetensors" size=6776MB
13:45:59-574446 INFO Load embeddings: loaded=0 skipped=1 time=0.02
13:45:59-731795 INFO Load model: time=3.30 load=3.30 native=1024 {'ram':
{'used': 1.14, 'total': 31.85}, 'gpu': {'used': 8.08,
'total': 15.99}, 'retries': 0, 'oom': 0}
13:45:59-731795 INFO Startup time: 10.81 torch=2.70 gradio=0.64
libraries=1.34 extensions=0.53 ui-en=0.14
ui-control=0.07 ui-settings=0.15 ui-extensions=1.02
ui-defaults=0.05 launch=0.10 api=0.05 app-started=0.28
checkpoint=3.47
13:46:21-076969 INFO Base: class=StableDiffusionXLPipeline
Progress 1.90s/it █████████████████████████████████ 100% 30/30 00:57 00:00 Base

[vladmandic]

re: sd upscale - that means that those scripts need to be updated to actually set job states correctly. i can do that for sd upscale, but not ultimate upscale as that is external extension.

re: starts at 40% - no, that is simply a consequence of ui polling intervals, you can set them as high or as low as you want.

re: tiling performance - that's a separate topic. we can look into that, but lets do that separately - combining issues in a single filing never works.