stable-diffusion-webui
stable-diffusion-webui copied to clipboard
AUTOMATIC1111
[Bug]: SD XL model loading on AMD - Failed to create model quickly; will retry using slow method
Is there an existing issue for this?
- [X] I have searched the existing issues and checked the recent builds/commits
What happened?
Trying to load the sd_xl_base_0.9.safetensors model ends with:
Creating model from config: /dockerx/repositories/generative-models/configs/inference/sd_xl_base.yaml Failed to create model quickly; will retry using slow method.
The file is in the folder present and should be readable.
Inside an docker container on a linux maschine with an AMD card. Maybe i missed a step.
Steps to reproduce the problem
Load sd_xl_base_0.9.safetensors via UI.
What should have happened?
Model should be loaded
Version or Commit where the problem happens
14cf434bc36d0ef31f31d4c6cd2bd15d7857d5c8
What Python version are you running on ?
Python 3.9.x (below, no recommended)
What platforms do you use to access the UI ?
Linux
What device are you running WebUI on?
AMD GPUs (RX 6000 above)
Cross attention optimization
Automatic
What browsers do you use to access the UI ?
Google Chrome
Command Line Arguments
TORCH_COMMAND='pip install torch==1.13.1+rocm5.2 torchvision==0.14.1+rocm5.2 --index-url https://download.pytorch.org/whl/rocm5.2' python launch.py --precision full --no-half --lyco-dir /dockerx/models/lycoris/
List of extensions
| PBRemTools | https://github.com/mattyamonaca/PBRemTools.git | main | d0fee2a8 | Thu Jun 29 14:08:23 2023 | latest |
|---|---|---|---|---|---|
| Stable-Diffusion-Webui-Civitai-Helper | https://github.com/butaixianran/Stable-Diffusion-Webui-Civitai-Helper.git | main | 920ca326 | Tue May 23 11:53:22 2023 | latest |
| a1111-sd-webui-lycoris | https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris.git | main | 8e97bf54 | Sun Jul 9 07:44:58 2023 | latest |
| a1111-sd-webui-tagcomplete | https://github.com/DominikDoom/a1111-sd-webui-tagcomplete.git | main | 737b6973 | Sat Jul 8 16:03:44 2023 | latest |
| model_preset_manager | https://github.com/rifeWithKaiju/model_preset_manager.git | main | 4e25eebd | Tue May 30 15:07:37 2023 | latest |
| sd-webui-controlnet | https://github.com/Mikubill/sd-webui-controlnet.git | main | 07bed6cc | Sat Jul 8 20:54:01 2023 | latest |
| stable-diffusion-webui-eyemask | https://github.com/ilian6806/stable-diffusion-webui-eyemask.git | master | 7b803a43 | Fri Jun 2 11:15:19 2023 | latest |
PBRemTools https://github.com/mattyamonaca/PBRemTools.git main d0fee2a8 Thu Jun 29 14:08:23 2023 latest Stable-Diffusion-Webui-Civitai-Helper https://github.com/butaixianran/Stable-Diffusion-Webui-Civitai-Helper.git main 920ca326 Tue May 23 11:53:22 2023 latest a1111-sd-webui-lycoris https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris.git main 8e97bf54 Sun Jul 9 07:44:58 2023 latest a1111-sd-webui-tagcomplete https://github.com/DominikDoom/a1111-sd-webui-tagcomplete.git main 737b6973 Sat Jul 8 16:03:44 2023 latest model_preset_manager https://github.com/rifeWithKaiju/model_preset_manager.git main 4e25eebd Tue May 30 15:07:37 2023 latest sd-webui-controlnet https://github.com/Mikubill/sd-webui-controlnet.git main 07bed6cc Sat Jul 8 20:54:01 2023 latest stable-diffusion-webui-eyemask https://github.com/ilian6806/stable-diffusion-webui-eyemask.git master 7b803a43 Fri Jun 2 11:15:19 2023 latest
Console logs
> Creating model from config: /dockerx/repositories/generative-models/configs/inference/sd_xl_base.yaml
> Failed to create model quickly; will retry using slow method.
Additional information
Its running in the following docker container image:
rocm/pytorch:rocm5.5_ubuntu20.04_py3.8_pytorch_1.13.1
Well for gfx906 I get this error
Loading weights [13406f993c] from /media/bcansin/ai/ai/stable-diffusion-webui/models/Stable-diffusion/sd_xl_refiner_0.9/sd_xl_refiner_0.9.safetensors
Creating model from config: /media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/configs/inference/sd_xl_refiner.yaml
Failed to create model quickly; will retry using slow method.
changing setting sd_model_checkpoint to sd_xl_refiner_0.9/sd_xl_refiner_0.9.safetensors [13406f993c]: AssertionError
Traceback (most recent call last):
File "/media/bcansin/ai/ai/stable-diffusion-webui/modules/shared.py", line 631, in set
self.data_labels[key].onchange()
File "/media/bcansin/ai/ai/stable-diffusion-webui/modules/call_queue.py", line 14, in f
res = func(*args, **kwargs)
File "/media/bcansin/ai/ai/stable-diffusion-webui/webui.py", line 238, in <lambda>
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights()), call=False)
File "/media/bcansin/ai/ai/stable-diffusion-webui/modules/sd_models.py", line 578, in reload_model_weights
load_model(checkpoint_info, already_loaded_state_dict=state_dict)
File "/media/bcansin/ai/ai/stable-diffusion-webui/modules/sd_models.py", line 504, in load_model
sd_model = instantiate_from_config(sd_config.model)
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/util.py", line 89, in instantiate_from_config
return get_obj_from_str(config["target"])(**config.get("params", dict()))
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/models/diffusion.py", line 65, in __init__
self._init_first_stage(first_stage_config)
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/models/diffusion.py", line 106, in _init_first_stage
model = instantiate_from_config(config).eval()
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/util.py", line 175, in instantiate_from_config
return get_obj_from_str(config["target"])(**config.get("params", dict()))
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/models/autoencoder.py", line 295, in __init__
self.encoder = Encoder(**ddconfig)
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/modules/diffusionmodules/model.py", line 553, in __init__
self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
File "/media/bcansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/modules/diffusionmodules/model.py", line 286, in make_attn
assert XFORMERS_IS_AVAILABLE, (
AssertionError: We do not support vanilla attention in 1.13.1+rocm5.2 anymore, as it is too expensive. Please install xformers via e.g. 'pip install xformers==0.0.16'
and also if I try with newest dev or stable one of pytorch this happend well I guess I really made big mistake to buy AMD Radeon VII https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/10873 And also heres pytorch/rocm offical github no reply yet... so yeah if you got AMD you can't. https://github.com/pytorch/pytorch/issues/103973 https://github.com/RadeonOpenCompute/ROCm/issues/2314
Is there an existing issue for this?
- [x] I have searched the existing issues and checked the recent builds/commits
What happened?
Trying to load the sd_xl_base_0.9.safetensors model ends with:
Creating model from config: /dockerx/repositories/generative-models/configs/inference/sd_xl_base.yaml Failed to create model quickly; will retry using slow method.
The file is in the folder present and should be readable.
Inside an docker container on a linux maschine with an AMD card. Maybe i missed a step.
Steps to reproduce the problem
Load sd_xl_base_0.9.safetensors via UI.
What should have happened?
Model should be loaded
Version or Commit where the problem happens
What Python version are you running on ?
Python 3.9.x (below, no recommended)
What platforms do you use to access the UI ?
Linux
What device are you running WebUI on?
AMD GPUs (RX 6000 above)
Cross attention optimization
Automatic
What browsers do you use to access the UI ?
Google Chrome
Command Line Arguments
TORCH_COMMAND='pip install torch==1.13.1+rocm5.2 torchvision==0.14.1+rocm5.2 --index-url https://download.pytorch.org/whl/rocm5.2' python launch.py --precision full --no-half --lyco-dir /dockerx/models/lycoris/List of extensions
PBRemTools https://github.com/mattyamonaca/PBRemTools.git main d0fee2a8 Thu Jun 29 14:08:23 2023 latest Stable-Diffusion-Webui-Civitai-Helper https://github.com/butaixianran/Stable-Diffusion-Webui-Civitai-Helper.git main 920ca326 Tue May 23 11:53:22 2023 latest a1111-sd-webui-lycoris https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris.git main 8e97bf54 Sun Jul 9 07:44:58 2023 latest a1111-sd-webui-tagcomplete https://github.com/DominikDoom/a1111-sd-webui-tagcomplete.git main 737b6973 Sat Jul 8 16:03:44 2023 latest model_preset_manager https://github.com/rifeWithKaiju/model_preset_manager.git main 4e25eebd Tue May 30 15:07:37 2023 latest sd-webui-controlnet https://github.com/Mikubill/sd-webui-controlnet.git main 07bed6cc Sat Jul 8 20:54:01 2023 latest stable-diffusion-webui-eyemask https://github.com/ilian6806/stable-diffusion-webui-eyemask.git master 7b803a43 Fri Jun 2 11:15:19 2023 latest PBRemTools https://github.com/mattyamonaca/PBRemTools.git main d0fee2a8 Thu Jun 29 14:08:23 2023 latest Stable-Diffusion-Webui-Civitai-Helper https://github.com/butaixianran/Stable-Diffusion-Webui-Civitai-Helper.git main 920ca326 Tue May 23 11:53:22 2023 latest a1111-sd-webui-lycoris https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris.git main 8e97bf54 Sun Jul 9 07:44:58 2023 latest a1111-sd-webui-tagcomplete https://github.com/DominikDoom/a1111-sd-webui-tagcomplete.git main 737b6973 Sat Jul 8 16:03:44 2023 latest model_preset_manager https://github.com/rifeWithKaiju/model_preset_manager.git main 4e25eebd Tue May 30 15:07:37 2023 latest sd-webui-controlnet https://github.com/Mikubill/sd-webui-controlnet.git main 07bed6cc Sat Jul 8 20:54:01 2023 latest stable-diffusion-webui-eyemask https://github.com/ilian6806/stable-diffusion-webui-eyemask.git master 7b803a43 Fri Jun 2 11:15:19 2023 latest
Console logs
> Creating model from config: /dockerx/repositories/generative-models/configs/inference/sd_xl_base.yaml > Failed to create model quickly; will retry using slow method.Additional information
Its running in the following docker container image:
rocm/pytorch:rocm5.5_ubuntu20.04_py3.8_pytorch_1.13.1
Try to load Fp16 model version, and use Fp16 VAE too. I think "Failed to create model quickly; will retry using slow method." error is related to not enough VRAM
I'm having the same issue with a Nvidia RTX 2070 Max-Q so it's not an AMD issue
I'm having the same issue with a Nvidia RTX 2070 Max-Q so it's not an AMD issue
For my test it looks so much like pytorch version my card won't support 2.0.1 but if I run with it it can open model but thx to rocm buggy for gfx906 and ending support in next release I guess I can't never generate at pytorch 2 or more newer.
I'm having the same issue under Ubuntu 22.04 with a Nvidia RTW 3090 when trying to load SDXL.
After getting the Failed to create model quickly; will retry using slow method it takes all fills up all my RAM.
I tried to leave it for 10 minutes or so at 100% RAM but nothing happened.
EDIT: I deleted the stable-diffusion-webui directory, cloned a new one, started webui.sh, let it reinstall everything and now it works.I'm guessing an outdated dependency or something like that (I already was on the last AUTOMATIC1111/stable-diffusion-webui version when I got the error by the way).
I'm getting same problem, also it downloading 10G file and my internet is suck with cmd, i need to download the file manually but where to put it ? any idea ?
Loading weights [31e35c80fc] from D:\stable-diffusion-webui\models\Stable-diffusion\sd_xl_base_1.0.safetensors2.89it/s] Creating model from config: D:\stable-diffusion-webui\repositories\generative-models\configs\inference\sd_xl_base.yaml Failed to create model quickly; will retry using slow method. Downloading (…)ip_pytorch_model.bin: 6%|██▍ | 619M/10.2G
I fixed this issue in my local virtual environment by running the webui-user script with --reinstall-torch commandline arg, then I ran it again with --xformers commandline arg. This forces the webui.py script to reinstall or update these modules and now I'm able to use both normal SDXL and DreamShaper XL.
I fixed this issue in my local virtual environment by running the webui-user script with
--reinstall-torchcommandline arg, then I ran it again with--xformerscommandline arg. This forces the webui.py script to reinstall or update these modules and now I'm able to use both normal SDXL and DreamShaper
How Much VRAM you Have?
Should i Run with this command like this:
@echo off
set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --reinstall-torch
call webui.bat
I have a 6GB Gtx 1660 SUPER GPU and 8 GB RAM and SD works fine for me except for this new model loading problem:
File "F:\Stable Diffusion Final\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\conv.py", line 137, in __init__
self.weight = Parameter(torch.empty(
RuntimeError: [enforce fail at ..\c10\core\impl\alloc_cpu.cpp:72] data. DefaultCPUAllocator: not enough memory: you tried to allocate 58982400 bytes.
Stable diffusion model failed to load
I have Tried All the methods to getrid of this error but i cant and let me tell you i have one Old model which is 10Gb and i can load that but i cant load this 6Gb Model sd_xl LOL
How Much VRAM you Have?
Should i Run with this command like this:
@echo off set PYTHON= set GIT= set VENV_DIR= set COMMANDLINE_ARGS=--xformers --reinstall-torch call webui.bat
I have 8GB RTX 3060.
Also when I tried running with COMMANDLINE_ARGS=--xformers --reinstall-torch it didn't work, I had to run COMMANDLINE_ARGS=--reinstall-torch, and then run again with COMMANDLINE_ARGS=--xformers, and I have no idea why it only worked that way.
You could also try using --lowram &--lowvramor --medram & --medvram commandline args, but I'm not sure if the XL models support them.
Okey I managed generate SDXL 1 or Dream SDXL or etc..
with invoke ai v3.0.1rc3 with
pip install torch==1.13.1+rocm5.2 torchvision==0.14.1+rocm5.2 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/rocm5.2
[2023-07-31 04:23:44,153]::[InvokeAI]::INFO --> Converting /media/b_cansin/ai/ai/stable-diffusion-webui/models/Stable-diffusion/dreamshaperXL10_alpha2Xl10/dreamshaperXL10_alpha2Xl10.safetensors to diffusers format
[2023-07-31 04:31:29,052]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:tokenizer
[2023-07-31 04:31:31,240]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:text_encoder
[2023-07-31 04:31:44,742]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:tokenizer_2
[2023-07-31 04:31:45,297]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:text_encoder_2
[2023-07-31 04:33:54,833]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:scheduler
[2023-07-31 04:33:55,596]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:unet
93%|████████████████████████████████████████████████████████████████████████████████████▉ | 140/150 [04:51<00:20, 2.08s/it]
[2023-07-31 04:40:00,087]::[InvokeAI]::INFO --> Loading model /media/b_cansin/ai/ai/InvokeAI-Installer/models/.cache/31f8e9c5a9debddf84f4c2a4bb8758aa, type sdxl:main:vae
And still that version can generate but for AUTOMATIC1111
version: [1.5.1](https://github.com/AUTOMATIC1111/stable-diffusion-webui/commit/7ba3923d5b494b7756d0b12f33acb3716d830b9a) • python: 3.10.6 • torch: 1.13.1+rocm5.2 • xformers: N/A • gradio: 3.32.0 • checkpoint: [879db523c3](https://google.com/search?q=879db523c30d3b9017143d56705015e15a2cb5628762c11d086fed9538abd7fd)
Calculating sha256 for /media/b_cansin/ai/ai/stable-diffusion-webui/models/Stable-diffusion/animeArtDiffusionXL_alpha3/animeArtDiffusionXL_alpha3.safetensors: 53bb4fdc63b36014201f2789eab73f3b2b3569a2a9a57b3efb28d4f17283e4c4
Loading weights [53bb4fdc63] from /media/b_cansin/ai/ai/stable-diffusion-webui/models/Stable-diffusion/animeArtDiffusionXL_alpha3/animeArtDiffusionXL_alpha3.safetensors
Creating model from config: /media/b_cansin/ai/ai/stable-diffusion-webui/repositories/generative-models/configs/inference/sd_xl_base.yaml
creating model quickly: AssertionError
Traceback (most recent call last):
File "/usr/lib/python3.10/threading.py", line 973, in _bootstrap
self._bootstrap_inner()
File "/usr/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
self.run()
File "/media/b_cansin/ai/ai/stable-diffusion-webui/venv/lib/python3.10/site-packages/anyio/_backends/_asyncio.py", line 807, in run
result = context.run(func, *args)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/modules/ui_settings.py", line 272, in <lambda>
fn=lambda value, k=k: self.run_settings_single(value, key=k),
File "/media/b_cansin/ai/ai/stable-diffusion-webui/modules/ui_settings.py", line 90, in run_settings_single
if not opts.set(key, value):
File "/media/b_cansin/ai/ai/stable-diffusion-webui/modules/shared.py", line 633, in set
self.data_labels[key].onchange()
File "/media/b_cansin/ai/ai/stable-diffusion-webui/modules/call_queue.py", line 14, in f
res = func(*args, **kwargs)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/webui.py", line 238, in <lambda>
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights()), call=False)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/modules/sd_models.py", line 582, in reload_model_weights
load_model(checkpoint_info, already_loaded_state_dict=state_dict)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/modules/sd_models.py", line 498, in load_model
sd_model = instantiate_from_config(sd_config.model)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/util.py", line 89, in instantiate_from_config
return get_obj_from_str(config["target"])(**config.get("params", dict()))
File "/media/b_cansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/models/diffusion.py", line 65, in __init__
self._init_first_stage(first_stage_config)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/models/diffusion.py", line 106, in _init_first_stage
model = instantiate_from_config(config).eval()
File "/media/b_cansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/util.py", line 175, in instantiate_from_config
return get_obj_from_str(config["target"])(**config.get("params", dict()))
File "/media/b_cansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/models/autoencoder.py", line 295, in __init__
self.encoder = Encoder(**ddconfig)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/repositories/generative-models/sgm/modules/diffusionmodules/model.py", line 553, in __init__
self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
File "/media/b_cansin/ai/ai/stable-diffusion-webui/
repositories/generative-models/sgm/modules/diffusionmodules/model.py", line 286, in make_attn
assert XFORMERS_IS_AVAILABLE, (
AssertionError: We do not support vanilla attention in 1.13.1+rocm5.2 anymore, as it is too expensive. Please install xformers via e.g. 'pip install xformers==0.0.16'
Thats getting intersting also before this I fight with Segmentation fault (core dumped) even crash python for only happend on AUTOMATIC1111 even other AIs working.
@tangcan1600 how did you solve this ?
I still have this issue sdwebui works with sd 1.5 but not with sdxl
stablediff-rocm-runner | Loading weights [31e35c80fc] from /stablediff-web/models/Stable-diffusion/Stable-diffusion/sdxl/sd_xl_base_1.0.safetensors
stablediff-rocm-runner | Creating model from config: /stablediff-web/repositories/generative-models/configs/inference/sd_xl_base.yaml
stablediff-rocm-runner | Failed to create model quickly; will retry using slow method.
Memory usage peaked as soon as the SDXL model was loaded.
I was using GPU 12GB VRAM RTX 3060. The journey with SD1.5 has been pleasant for the last few months. I ran into a problem with SDXL not loading properly in Automatic1111 Version 1.5. The characteristic situation was severe system-wide stuttering that I never experienced before. I was using 8gbX2 of dual channel system memory, so I just upgraded to 16gbx2. Immediately after the upgrade, I loaded the SDXL base model through the same process and confirmed that it loaded successfully. I don't think this is the only solution, but it is one of the physically possible solutions.
Solution:
- Create the path "openai" under the path "stable-diffusion" (whichever is your path name).
- Use git to clone the repository under the "openai" path: "https://www.modelscope.cn/AI-ModelScope/clip-vit-large-patch14. git".
- Once done, restart webui.sh or webui.bat(windows pc) under stable-diffusion.
- This issue is not related to Nvidia or AMD graphics cards.
- You'll notice that your pages using the nginx reverse proxy are also extremely smooth!
I also encountered this problem. I was about to give up and restarted the computer. Then I only used --xformers and re-ran run.bat and it worked normally. I hope this can help you.
