Cannot get UI to load (after switching model)

[raffetazarius]

I had a fresh installation of SD.Next running fine on my MacBook Pro M3 Max 128GB 8TB.

The only thing I did was switched the model, and it seems to crash the UI and I cannot get it load again.

I've tried loading using both the Original and Diffusers backends (using the --backend arg), but no luck.

Is there any way I can get this going again without doing another reinstall?

---

privateuser@Users-MBP automatic % ./webui.sh --disable-queue
Create and activate python venv
Launching launch.py...
19:20:44-133645 INFO Starting SD.Next
19:20:44-135242 INFO Logger: file="/Users/privateuser/automatic/sdnext.log" level=INFO size=19863 mode=append
19:20:44-135765 INFO Python 3.10.13 on Darwin
19:20:44-154780 INFO Version: app=sd.next updated=2024-05-07 hash=e081f232 branch=master url=https://github.com/vladmandic/automatic/tree/master
19:20:44-570198 INFO Platform: arch=arm64 cpu=arm system=Darwin release=23.4.0 python=3.10.13
19:20:44-597896 WARNING Modified files: ['models/Reference/DeepFloyd--IF-I-M-v1.0.jpg', 'models/Reference/PixArt-alpha--PixArt-XL-2-1024-MS.jpg',
'models/Reference/PixArt-alpha--PixArt-XL-2-512x512.jpg', 'models/Reference/SimianLuo--LCM_Dreamshaper_v7.jpg',
'models/Reference/TempestV0.1-Artistic.jpg', 'models/Reference/XCLiu--instaflow_0_9B_from_sd_1_5.jpg',
'models/Reference/amused--amused-256.jpg', 'models/Reference/amused--amused-512.jpg',
'models/Reference/damo-vilab--text-to-video-ms-1.7b.jpg', 'models/Reference/dreamshaperXL_v21TurboDPMSDE.jpg',
'models/Reference/dreamshaper_8.jpg', 'models/Reference/etri-vilab--koala-700m-llava-cap.jpg',
'models/Reference/juggernautXL_v9Rdphoto2Lightning.jpg', 'models/Reference/juggernautXL_v9Rundiffusionphoto2.jpg',
'models/Reference/juggernaut_reborn.jpg', 'models/Reference/kandinsky-community--kandinsky-2-1.jpg',
'models/Reference/kandinsky-community--kandinsky-2-2-decoder.jpg', 'models/Reference/kandinsky-community--kandinsky-3.jpg',
'models/Reference/latent-consistency--lcm-sdxl.jpg', 'models/Reference/playgroundai--playground-v1.jpg',
'models/Reference/playgroundai--playground-v2-1024px-aesthetic.jpg', 'models/Reference/playgroundai--playground-v2-256px-base.jpg',
'models/Reference/playgroundai--playground-v2-512px-base.jpg', 'models/Reference/playgroundai--playground-v2.5-1024px-aesthetic.jpg',
'models/Reference/salesforce--blipdiffusion.jpg', 'models/Reference/sd_xl_base_1.0.jpg', 'models/Reference/sdxl_turbo.jpg',
'models/Reference/segmind--SSD-1B.jpg', 'models/Reference/segmind--SegMoE-2x1-v0.jpg', 'models/Reference/segmind--SegMoE-4x2-v0.jpg',
'models/Reference/segmind--SegMoE-SD-4x2-v0.jpg', 'models/Reference/segmind--Segmind-Vega.jpg',
'models/Reference/segmind--tiny-sd.jpg', 'models/Reference/stabilityai--stable-cascade.jpg',
'models/Reference/stabilityai--stable-diffusion-2-1-base.jpg', 'models/Reference/stabilityai--stable-diffusion-2-1.jpg',
'models/Reference/stabilityai--stable-video-diffusion-img2vid-xt.jpg',
'models/Reference/stabilityai--stable-video-diffusion-img2vid.jpg', 'models/Reference/thu-ml--unidiffuser-v1.jpg',
'models/Reference/v1-5-pruned-fp16-emaonly.jpg', 'models/Reference/warp-ai--wuerstchen.jpg', 'models/VAE-approx/model.pt',
'models/karlo/ViT-L-14_stats.th']
19:20:44-610200 INFO Extensions: disabled=[]
19:20:44-611247 INFO Extensions: enabled=['sd-extension-chainner', 'sd-webui-controlnet', 'stable-diffusion-webui-rembg', 'sd-webui-agent-scheduler',
'Lora', 'sd-extension-system-info', 'stable-diffusion-webui-images-browser'] extensions-builtin
19:20:44-612694 INFO Extensions: enabled=[] extensions
19:20:44-613287 INFO Startup: quick launch
19:20:44-613736 INFO Verifying requirements
19:20:44-622136 INFO Verifying packages
19:20:44-623200 INFO Extensions: disabled=[]
19:20:44-623680 INFO Extensions: enabled=['sd-extension-chainner', 'sd-webui-controlnet', 'stable-diffusion-webui-rembg', 'sd-webui-agent-scheduler',
'Lora', 'sd-extension-system-info', 'stable-diffusion-webui-images-browser'] extensions-builtin
19:20:44-624717 INFO Extensions: enabled=[] extensions
19:20:44-627119 INFO Command line args: ['--disable-queue'] disable_queue=True
19:20:46-642281 INFO Load packages: {'torch': '2.3.0', 'diffusers': '0.27.0', 'gradio': '3.43.2'}
19:20:47-113992 INFO Engine: backend=Backend.DIFFUSERS compute=mps device=mps attention="Scaled-Dot-Product" mode=no_grad
19:20:47-116001 INFO Device:
19:20:47-410838 INFO Available VAEs: path="models/VAE" items=2
19:20:47-411494 INFO Disabled extensions: ['sd-webui-controlnet']
19:20:47-414419 INFO Available models: path="models/Stable-diffusion" items=55 time=0.00
19:20:47-453286 INFO LoRA networks: available=23 folders=2
19:20:47-681901 INFO Extension: script='extensions-builtin/sd-webui-agent-scheduler/scripts/task_scheduler.py' Using sqlite file:
extensions-builtin/sd-webui-agent-scheduler/task_scheduler.sqlite3
19:20:47-795134 INFO UI theme: name="black-teal" style=Auto base=sdnext.css
19:20:48-445895 INFO Extension list is empty: refresh required
19:20:48-686400 INFO Server queues disabled
19:20:48-734809 INFO Local URL: http://127.0.0.1:7860/
19:20:48-830533 INFO [AgentScheduler] Total pending tasks: 1
19:20:48-834660 INFO [AgentScheduler] Executing task task(xtr3jxmms1hezac)
19:20:48-835197 INFO [AgentScheduler] Registering APIs
19:20:49-013094 INFO Select: model="SD 1.5/absolutereality_v181 [463d6a9fe8]"
19:20:49-037990 INFO Setting Torch parameters: device=mps dtype=torch.float32 vae=torch.float32 unet=torch.float32 context=no_grad fp16=None bf16=None
optimization=Scaled-Dot-Product
19:20:49-045087 INFO Autodetect: model="Stable Diffusion" class=StableDiffusionPipeline file="/Users/privateuser/automatic/models/Stable-diffusion/SD
1.5/absolutereality_v181.safetensors" size=2034MB
19:20:49-084150 INFO Startup time: 4.41 torch=1.64 gradio=0.34 libraries=0.76 extensions=0.36 ui-en=0.08 ui-control=0.05 ui-settings=0.11
ui-extensions=0.48 ui-defaults=0.06 app-started=0.26
19:20:51-033540 INFO Load embeddings: loaded=14 skipped=11 time=0.16
19:20:51-158657 INFO Load model: time=2.01 load=2.00 native=512 {'ram': {'used': 4.07, 'total': 128.0}}
19:20:51-161068 INFO Select: model="SD 1.5/cirrostratus_v25 [9abd72c940]"
19:20:51-672081 INFO Setting Torch parameters: device=mps dtype=torch.float32 vae=torch.float32 unet=torch.float32 context=no_grad fp16=None bf16=None
optimization=Scaled-Dot-Product
19:20:51-672862 INFO Autodetect: model="Stable Diffusion" class=StableDiffusionPipeline file="/Users/privateuser/automatic/models/Stable-diffusion/SD
1.5/cirrostratus_v25.safetensors" size=2193MB
19:20:53-089710 INFO Load embeddings: loaded=14 skipped=11 time=0.14
19:20:53-212810 INFO Load model: time=1.42 load=1.42 native=512 {'ram': {'used': 4.31, 'total': 128.0}}
19:20:53-224745 INFO Base: class=StableDiffusionImg2ImgPipeline
Progress ?it/s 0% 0/20 00:00 ? Base/AppleInternal/Library/BuildRoots/ce725a5f-c761-11ee-a4ec-b6ef2fd8d87b/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShaders/MPSCore/Types/MPSNDArray.mm:788: failed assertion `[MPSNDArray initWithDevice:descriptor:] Error: total bytes of NDArray > 2**32'
zsh: abort ./webui.sh --disable-queue
privateuser@Users-MBP automatic % /Users/privateuser/.pyenv/versions/3.10.13/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '

[vladmandic]

model cirrostratus_v25 you've tried has a really strange size for sd15 model and seems like apple mps has a limit to max size and it crashes.
but when you restart sdnext, it tries to load same model again.
i suggest to either edit config json to remove that model so its no longer auto-loaded on startup and pick some normal model.

[raffetazarius]

Thanks Vlad!

I've removed reference to the model from config.json, but the same crash occurs trying to start up SD.Next:

07:25:11-752099 INFO Base: class=StableDiffusionImg2ImgPipeline
Progress ?it/s 0% 0/20 00:00 ? Base/AppleInternal/Library/BuildRoots/ce725a5f-c761-11ee-a4ec-b6ef2fd8d87b/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShaders/MPSCore/Types/MPSNDArray.mm:788: failed assertion `[MPSNDArray initWithDevice:descriptor:] Error: total bytes of NDArray > 2**32'
zsh: abort ./webui.sh
privateusers@Users-MBP automatic % /Users/privateuser/.pyenv/versions/3.10.13/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '
privateuser@Users-MBP automatic %

Is there a way to force SD.Next not to load the model on start up, using a cmd arg? Then I can get into the UI to disable auto-loading of a model on startup (so that this doesn't occur again and a restart of the UI actually restores functionality as it should)

[raffetazarius]

I've also tried inserting a DIFFERENT model into line 37 of that file, it has no effect on the SD.Next startup sequence.

I've also tried deleting the cache.json file from the automatic folder (I believe this contains hashes of models).

Now a different error occurs and I still can't load SD.Next UI.

---

privateuser@Users-MBP automatic % ./webui.sh
Create and activate python venv
Launching launch.py...
07:46:12-164431 INFO Starting SD.Next
07:46:12-166010 INFO Logger:
file="/Users/privateuser/automatic/sdnext.log"
level=INFO size=96102 mode=append
07:46:12-166595 INFO Python 3.10.13 on Darwin
07:46:12-185682 INFO Version: app=sd.next updated=2024-05-07 hash=e081f232
branch=master
url=https://github.com/vladmandic/automatic/tree/master
07:46:12-549218 INFO Platform: arch=arm64 cpu=arm system=Darwin
release=23.4.0 python=3.10.13
07:46:12-582009 WARNING Modified files:
['models/Reference/DeepFloyd--IF-I-M-v1.0.jpg',
'models/Reference/PixArt-alpha--PixArt-XL-2-1024-MS.jpg
',
'models/Reference/PixArt-alpha--PixArt-XL-2-512x512.jpg
',
'models/Reference/SimianLuo--LCM_Dreamshaper_v7.jpg',
'models/Reference/TempestV0.1-Artistic.jpg',
'models/Reference/XCLiu--instaflow_0_9B_from_sd_1_5.jpg
', 'models/Reference/amused--amused-256.jpg',
'models/Reference/amused--amused-512.jpg',
'models/Reference/damo-vilab--text-to-video-ms-1.7b.jpg
', 'models/Reference/dreamshaperXL_v21TurboDPMSDE.jpg',
'models/Reference/dreamshaper_8.jpg',
'models/Reference/etri-vilab--koala-700m-llava-cap.jpg'
,
'models/Reference/juggernautXL_v9Rdphoto2Lightning.jpg'
,
'models/Reference/juggernautXL_v9Rundiffusionphoto2.jpg
', 'models/Reference/juggernaut_reborn.jpg',
'models/Reference/kandinsky-community--kandinsky-2-1.jp
g',
'models/Reference/kandinsky-community--kandinsky-2-2-de
coder.jpg',
'models/Reference/kandinsky-community--kandinsky-3.jpg'
, 'models/Reference/latent-consistency--lcm-sdxl.jpg',
'models/Reference/playgroundai--playground-v1.jpg',
'models/Reference/playgroundai--playground-v2-1024px-ae
sthetic.jpg',
'models/Reference/playgroundai--playground-v2-256px-bas
e.jpg',
'models/Reference/playgroundai--playground-v2-512px-bas
e.jpg',
'models/Reference/playgroundai--playground-v2.5-1024px-
aesthetic.jpg',
'models/Reference/salesforce--blipdiffusion.jpg',
'models/Reference/sd_xl_base_1.0.jpg',
'models/Reference/sdxl_turbo.jpg',
'models/Reference/segmind--SSD-1B.jpg',
'models/Reference/segmind--SegMoE-2x1-v0.jpg',
'models/Reference/segmind--SegMoE-4x2-v0.jpg',
'models/Reference/segmind--SegMoE-SD-4x2-v0.jpg',
'models/Reference/segmind--Segmind-Vega.jpg',
'models/Reference/segmind--tiny-sd.jpg',
'models/Reference/stabilityai--stable-cascade.jpg',
'models/Reference/stabilityai--stable-diffusion-2-1-bas
e.jpg',
'models/Reference/stabilityai--stable-diffusion-2-1.jpg
',
'models/Reference/stabilityai--stable-video-diffusion-i
mg2vid-xt.jpg',
'models/Reference/stabilityai--stable-video-diffusion-i
mg2vid.jpg',
'models/Reference/thu-ml--unidiffuser-v1.jpg',
'models/Reference/v1-5-pruned-fp16-emaonly.jpg',
'models/Reference/warp-ai--wuerstchen.jpg',
'models/VAE-approx/model.pt',
'models/karlo/ViT-L-14_stats.th']
07:46:12-593903 INFO Extensions: disabled=[]
07:46:12-594855 INFO Extensions: enabled=['sd-extension-chainner',
'sd-webui-controlnet', 'stable-diffusion-webui-rembg',
'sd-webui-agent-scheduler', 'Lora',
'sd-extension-system-info',
'stable-diffusion-webui-images-browser']
extensions-builtin
07:46:12-596227 INFO Extensions: enabled=[] extensions
07:46:12-596816 INFO Startup: quick launch
07:46:12-597258 INFO Verifying requirements
07:46:12-605600 INFO Verifying packages
07:46:12-606630 INFO Extensions: disabled=[]
07:46:12-607080 INFO Extensions: enabled=['sd-extension-chainner',
'sd-webui-controlnet', 'stable-diffusion-webui-rembg',
'sd-webui-agent-scheduler', 'Lora',
'sd-extension-system-info',
'stable-diffusion-webui-images-browser']
extensions-builtin
07:46:12-608019 INFO Extensions: enabled=[] extensions
07:46:12-610256 INFO Command line args: []
07:46:14-708523 INFO Load packages: {'torch': '2.3.0', 'diffusers': '0.27.0', 'gradio': '3.43.2'}
07:46:15-244759 INFO Engine: backend=Backend.DIFFUSERS compute=mps device=mps attention="Scaled-Dot-Product" mode=no_grad
07:46:15-246964 INFO Device:
07:46:15-532723 INFO Available VAEs: path="models/VAE" items=2
07:46:15-533375 INFO Disabled extensions: ['sd-webui-controlnet']
07:46:15-536210 INFO Available models: path="models/Stable-diffusion" items=55 time=0.00
07:46:15-579827 INFO LoRA networks: available=23 folders=2
07:46:15-812415 INFO Extension: script='extensions-builtin/sd-webui-agent-scheduler/scripts/task_scheduler.py' Using sqlite file:
extensions-builtin/sd-webui-agent-scheduler/task_scheduler.sqlite3
07:46:15-925766 INFO UI theme: name="black-teal" style=Auto base=sdnext.css
07:46:16-562662 INFO Extension list is empty: refresh required
07:46:16-883937 INFO Local URL: http://127.0.0.1:7860/
07:46:17-104495 INFO [AgentScheduler] Total pending tasks: 1
07:46:17-108491 INFO [AgentScheduler] Executing task task(xtr3jxmms1hezac)
07:46:17-109000 INFO [AgentScheduler] Registering APIs
07:46:17-192291 INFO Select: model="SD Turbo/sd_turbo"
07:46:17-193643 INFO Setting Torch parameters: device=mps dtype=torch.float32 vae=torch.float32 unet=torch.float32 context=no_grad fp16=None bf16=None optimization=Scaled-Dot-Product
07:46:17-194548 INFO Startup time: 4.58 torch=1.71 gradio=0.36 libraries=0.81 extensions=0.37 ui-en=0.08 ui-control=0.05 ui-settings=0.11 ui-extensions=0.48 ui-defaults=0.06
launch=0.08 app-started=0.26
07:46:17-196464 INFO Autodetect: model="Stable Diffusion 2" class=StableDiffusionPipeline file="/Users/privateuser/automatic/models/Stable-diffusion/SD Turbo/sd_turbo.safetensors"
size=4973MB
text_encoder/config.json: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 633/633 [00:00<00:00, 681kB/s]
tokenizer/tokenizer_config.json: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 824/824 [00:00<00:00, 3.01MB/s]
tokenizer/vocab.json: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1.06M/1.06M [00:00<00:00, 4.13MB/s]
tokenizer/merges.txt: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 525k/525k [00:00<00:00, 43.2MB/s]
tokenizer/special_tokens_map.json: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 460/460 [00:00<00:00, 1.46MB/s]
Calculating hash: /Users/privateuser/automatic/models/Stable-diffusion/SD Turbo/sd_turbo.safetensors ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.2/5.2 GB 0:00:00
07:46:22-516089 ERROR Model move: device=mps Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta
to a different device.
07:46:22-540593 INFO Load embeddings: loaded=0 skipped=25 time=0.02
07:46:22-674658 INFO Load model: time=5.35 load=5.35 native=768 {'ram': {'used': 0.78, 'total': 128.0}}
07:46:22-677179 WARNING Override not found: checkpoint=SD 1.5/cirrostratus_v25 [9abd72c940]
07:46:22-677665 INFO Select: model="SD Turbo/sd_turbo [3f067a1b94]"
07:46:22-691636 ERROR Model move: device=mps Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta
to a different device.
07:46:22-692249 INFO Base: class=StableDiffusionImg2ImgPipeline
07:46:22-765062 ERROR Prompt parser encode: Tensor on device mps:0 is not on the expected device meta!
07:46:22-774266 ERROR Model move: device=mps Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta
to a different device.
07:46:22-824124 ERROR Processing: args={'prompt': [''], 'negative_prompt': [''], 'guidance_scale': 12, 'generator': [<torch._C.Generator object at 0x3acd77a70>],
'callback_on_step_end': <function process_diffusers..diffusers_callback at 0x35b9e1900>, 'callback_on_step_end_tensor_inputs': ['latents',
'prompt_embeds', 'negative_prompt_embeds'], 'num_inference_steps': 41, 'eta': 1.0, 'output_type': 'latent', 'image': [<PIL.Image.Image image mode=RGB
size=1024x1536 at 0x38D253520>], 'strength': 0.5} Cannot copy out of meta tensor; no data!
07:46:22-825485 ERROR Processing: NotImplementedError
╭─────────────────────────────────────────────────────────────────────────── Traceback (most recent call last) ───────────────────────────────────────────────────────────────────────────╮
│ /Users/privateuser/automatic/modules/processing_diffusers.py:426 in process_diffusers │
│ │
│ 425 │ │ sd_models.move_model(shared.sd_model, devices.device) │
│ ❱ 426 │ │ output = shared.sd_model(**base_args) # pylint: disable=not-callable │
│ 427 │ │ if isinstance(output, dict): │
│ │
│ /Users/privateuser/automatic/venv/lib/python3.10/site-packages/torch/utils/_contextlib.py:115 in decorate_context │
│ │
│ 114 │ │ with ctx_factory(): │
│ ❱ 115 │ │ │ return func(*args, **kwargs) │
│ 116 │
│ │
│ /Users/privateuser/automatic/venv/lib/python3.10/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py:980 in call │
│ │
│ 979 │ │ ) │
│ ❱ 980 │ │ prompt_embeds, negative_prompt_embeds = self.encode_prompt( │
│ 981 │ │ │ prompt, │
│ │
│ /Users/privateuser/automatic/venv/lib/python3.10/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py:443 in encode_prompt │
│ │
│ 442 │ │ │
│ ❱ 443 │ │ prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device) │
│ 444 │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
NotImplementedError: Cannot copy out of meta tensor; no data!
07:46:22-912771 INFO Processed: images=0 time=0.23 its=0.00 memory={'ram': {'used': 0.82, 'total': 128.0}}
07:46:22-919794 INFO Select: model="SD Turbo/sd_turbo [3f067a1b94]"
07:46:22-930149 INFO [AgentScheduler] Task queue is empty

[vladmandic]

try webui --ckpt none
and if that doesn't help, you can manually add "sd_checkpoint_autoload": false, to config.json