Closed mng4 closed 6 months ago
Disty0
commented
6 months ago Did you compile torchvision? We need torchvision too.
You can export TORCH_COMMAND environment variable to override the installer.
Default for CUDA is: TORCH_COMMAND="torch torchvision --index-url https://download.pytorch.org/whl/cu121"
mng4
commented
6 months ago @Disty0 Sorry my bad I missed it, I thought I only needed torch. I just compiled torchvision with v0.15.2 (assuming compatiable torch and torchvision as they were both released on same day.)
python -c "import torch; import torchvision; print(torch.__version__); print(torchvision.__version__);print(torch.randn(1).cuda())"
Response as
2.0.0a0+gite9ebda2 0.15.2a0+fa99a53 tensor([-1.2298], device='cuda:0')
Do I try to install sdnext once again?
Or Shall I run
pip install torch==2.0.1+cu113 torchvision==0.15.2+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
or
conda install pytorch==2.0.1 torchvision==0.15.2 cudatoolkit=11.3 -c pytorch
before trying to install SDNext?
vladmandic
commented
6 months ago if you're running manually compiled torch, perhaps best to tell sdnext to skip torch tests and simply try to use it as-is
just add --skip-torch cmd flag.
mng4
commented
6 months ago Thanks @vladmandic that's what I was just looking for.
mng4
commented
6 months ago Closing as the issue resolved.
mng4
commented
6 months ago Sorry to open the issue again, I am able to launch SD Next but not able to generate any image. It seems somehow SDNext still loading or trying to load torch 2.2.2
04:22:31-811817 INFO Verifying requirements
04:22:31-844392 DEBUG Setup complete without errors: 1712443952
04:22:31-862779 DEBUG Extension preload: {'extensions-builtin': 0.01, 'extensions': 0.0}
04:22:31-865861 DEBUG Starting module: <module 'webui' from 'webui.py'>
04:22:31-869066 INFO Command line args: ['--skip-torch', '--debug', '--lowvram', '--listen'] lowvram=True listen=True skip_torch=True debug=True
04:22:31-872374 DEBUG Env flags: []
04:22:48-042340 INFO Load packages: {'torch': '2.2.2+cu121', 'diffusers': '0.27.0', 'gradio': '3.43.2'}
04:22:51-860877 DEBUG Read: file="config.json" json=21 bytes=885 time=0.000
04:22:51-870605 INFO Engine: backend=Backend.DIFFUSERS compute=cuda device=cuda attention="Scaled-Dot-Product" mode=no_grad
04:22:51-876284 INFO Device: error=The NVIDIA driver on your system is too old (found version 11040). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx
Alternatively, go to: https://pytorch.org to install a PyTorch version that has been compiled with your version of the CUDA driver. Please let me know how do I disable this one.
vladmandic
commented
6 months ago do you see Skipping Torch tests message just before the log snipped you've posted (best to include entire startup log in the future)
anyhow, its not SDNext, its Pythons choice which torch is going to get used with import torch
for example - if your torch is installed in global context and SDNext already has torch installed in venv, python will prioritize which one?
you need to find torch-2.2.1 and remove it so only torch that is available for import is yours
mng4
commented
6 months ago Sorry my bad here is the complete log.
07:49:20-970508 INFO Starting SD.Next
07:49:20-980411 INFO Logger: file="/media/sinfinities/SDNext/sdnext.log" level=DEBUG size=64 mode=create
07:49:20-982736 INFO Python 3.9.19 on Linux
07:49:21-155071 INFO Version: app=sd.next updated=2024-04-06 hash=e783b098 branch=master url=https://github.com/vladmandic/automatic/tree/master
07:49:21-893411 INFO Platform: arch=x86_64 cpu= system=Linux release=6.1.0-19-amd64 python=3.9.19
07:49:21-896269 DEBUG Setting environment tuning
07:49:21-899370 DEBUG HF cache folder: /home/sinfinities/.cache/huggingface/hub
07:49:21-902409 INFO Skipping Torch tests
07:49:22-988905 DEBUG Repository update time: Sat Apr 6 06:30:23 2024
07:49:22-996619 INFO Startup: standard
07:49:23-002870 INFO Verifying requirements
07:49:23-040221 INFO Verifying packages
07:49:23-044635 INFO Verifying submodules
07:49:25-375014 DEBUG Submodule: extensions-builtin/sd-extension-chainner / main
07:49:25-565157 DEBUG Submodule: extensions-builtin/sd-extension-system-info / main
07:49:25-663235 DEBUG Submodule: extensions-builtin/sd-webui-agent-scheduler / main
07:49:25-775278 DEBUG Submodule: extensions-builtin/sd-webui-controlnet / main
07:49:26-154349 DEBUG Submodule: extensions-builtin/stable-diffusion-webui-images-browser / main
07:49:26-247255 DEBUG Submodule: extensions-builtin/stable-diffusion-webui-rembg / master
07:49:26-313330 DEBUG Submodule: modules/k-diffusion / master
07:49:26-421305 DEBUG Submodule: wiki / master
07:49:26-556949 DEBUG Register paths
07:49:27-012887 DEBUG Installed packages: 240
07:49:27-024641 DEBUG Extensions all: ['Lora', 'sd-extension-chainner', 'sd-extension-system-info', 'sd-webui-agent-scheduler', 'sd-webui-controlnet', 'stable-diffusion-webui-images-browser', 'stable-diffusion-webui-rembg']
07:49:27-911549 DEBUG Running extension installer: /media/sinfinities/SDNext/extensions-builtin/sd-extension-system-info/install.py
07:49:28-862529 DEBUG Running extension installer: /media/sinfinities/SDNext/extensions-builtin/sd-webui-agent-scheduler/install.py
07:49:29-800328 DEBUG Running extension installer: /media/sinfinities/SDNext/extensions-builtin/sd-webui-controlnet/install.py
07:49:30-978081 DEBUG Running extension installer: /media/sinfinities/SDNext/extensions-builtin/stable-diffusion-webui-images-browser/install.py
07:49:31-846900 DEBUG Running extension installer: /media/sinfinities/SDNext/extensions-builtin/stable-diffusion-webui-rembg/install.py
07:49:32-730769 DEBUG Extensions all: []
07:49:32-737925 INFO Extensions enabled: ['Lora', 'sd-extension-chainner', 'sd-extension-system-info', 'sd-webui-agent-scheduler', 'sd-webui-controlnet', 'stable-diffusion-webui-images-browser', 'stable-diffusion-webui-rembg']
07:49:32-745238 INFO Verifying requirements
07:49:32-783762 DEBUG Setup complete without errors: 1712542773
07:49:32-808476 DEBUG Extension preload: {'extensions-builtin': 0.02, 'extensions': 0.0}
07:49:32-815246 DEBUG Starting module: <module 'webui' from 'webui.py'>
07:49:32-823061 INFO Command line args: ['--skip-torch', '--debug', '--lowvram', '--listen', '--use-cuda'] lowvram=True listen=True use_cuda=True skip_torch=True debug=True
07:49:32-831658 DEBUG Env flags: []
07:49:48-745889 INFO Load packages: {'torch': '2.2.2+cu121', 'diffusers': '0.27.0', 'gradio': '3.43.2'}
07:49:52-692533 DEBUG Read: file="config.json" json=14 bytes=613 time=0.000
07:49:52-701479 INFO Engine: backend=Backend.DIFFUSERS compute=cuda device=cuda attention="Scaled-Dot-Product" mode=no_grad
07:49:52-707008 INFO Device: error=The NVIDIA driver on your system is too old (found version 11040). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx
Alternatively, go to: https://pytorch.org to install a PyTorch version that has been compiled with your version of the CUDA driver.
07:49:52-713030 DEBUG Migrated styles: file=styles.csv folder=models/styles
07:49:52-727801 DEBUG Load styles: folder="models/styles" items=288 time=0.01
07:49:52-732713 DEBUG Read: file="html/reference.json" json=36 bytes=21248 time=0.001
07:49:53-672420 DEBUG Importing LDM
07:49:53-804370 DEBUG Entering start sequence
07:49:53-820432 DEBUG Initializing
07:49:53-939418 INFO Available VAEs: path="models/VAE" items=0
07:49:53-946290 INFO Disabled extensions: ['sd-webui-controlnet']
07:49:53-958174 DEBUG Scanning diffusers cache: folder=models/Diffusers items=0 time=0.00
07:49:53-962579 INFO Available models: path="models/Stable-diffusion" items=0 time=0.01
07:49:56-822207 DEBUG Load extensions
07:49:57-858467 INFO LoRA networks: available=0 folders=2
07:50:03-246411 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
07:50:05-153286 INFO Extension: script='extensions-builtin/stable-diffusion-webui-images-browser/scripts/image_browser.py' Image Browser: Creating database
07:50:05-158865 INFO Extension: script='extensions-builtin/stable-diffusion-webui-images-browser/scripts/image_browser.py' Image Browser: Database created
07:50:05-381630 DEBUG Extensions init time: 8.56 animatediff.py=0.05 demofusion.py=0.08 differential_diffusion.py=0.09 loopback.py=0.08 Lora=0.29 sd-extension-chainner=0.99 sd-webui-agent-scheduler=4.31
stable-diffusion-webui-images-browser=1.91
07:50:05-730667 DEBUG Read: file="html/upscalers.json" json=4 bytes=2640 time=0.002
07:50:05-736646 INFO Upscaler create: folder="models/chaiNNer"
07:50:05-744618 DEBUG Read: file="extensions-builtin/sd-extension-chainner/models.json" json=24 bytes=2693 time=0.001
07:50:05-768582 DEBUG chaiNNer models: path="models/chaiNNer" defined=24 discovered=0 downloaded=0
07:50:05-785446 INFO Upscaler create: folder="models/RealESRGAN"
07:50:05-808234 DEBUG Load upscalers: total=52 downloaded=0 user=0 time=0.42 ['None', 'Lanczos', 'Nearest', 'ChaiNNer', 'ESRGAN', 'LDSR', 'RealESRGAN', 'SCUNet', 'SD', 'SwinIR']
07:50:05-837128 DEBUG Load styles: folder="models/styles" items=288 time=0.02
07:50:05-930394 DEBUG Creating UI
07:50:05-937326 INFO UI theme: name="black-teal" style=Auto base=sdnext.css
07:50:06-014030 DEBUG UI initialize: txt2img
07:50:06-100877 DEBUG Extra networks: page='model' items=36 subfolders=2 tab=txt2img folders=['models/Stable-diffusion', 'models/Diffusers', 'models/Reference'] list=0.03 thumb=0.01 desc=0.00 info=0.00 workers=4
07:50:06-112836 DEBUG Extra networks: page='style' items=288 subfolders=1 tab=txt2img folders=['models/styles', 'html'] list=0.01 thumb=0.00 desc=0.00 info=0.00 workers=4
07:50:06-117043 DEBUG Extra networks: page='embedding' items=0 subfolders=0 tab=txt2img folders=['models/embeddings'] list=0.00 thumb=0.00 desc=0.00 info=0.00 workers=4
07:50:06-120658 DEBUG Extra networks: page='hypernetwork' items=0 subfolders=0 tab=txt2img folders=['models/hypernetworks'] list=0.00 thumb=0.00 desc=0.00 info=0.00 workers=4
07:50:06-124982 DEBUG Extra networks: page='vae' items=0 subfolders=0 tab=txt2img folders=['models/VAE'] list=0.00 thumb=0.00 desc=0.00 info=0.00 workers=4
07:50:06-128761 DEBUG Extra networks: page='lora' items=0 subfolders=0 tab=txt2img folders=['models/Lora', 'models/LyCORIS'] list=0.00 thumb=0.00 desc=0.00 info=0.00 workers=4
07:50:06-246308 DEBUG UI initialize: img2img
07:50:06-398262 DEBUG UI initialize: control models=models/control
07:50:06-917502 ERROR Reading failed: ui-config.json [Errno 2] No such file or directory: 'ui-config.json'
07:50:07-337729 DEBUG Themes: builtin=12 gradio=5 huggingface=0
07:50:08-184132 INFO Extension list is empty: refresh required
07:50:08-852914 DEBUG Extension list: processed=7 installed=7 enabled=6 disabled=1 visible=7 hidden=0
07:50:08-976073 DEBUG Save: file="ui-config.json" json=0 bytes=2 time=0.006
07:50:09-290226 DEBUG Root paths: ['']
07:50:09-529215 INFO Local URL: http://localhost:7860/
07:50:09-535479 DEBUG Gradio functions: registered=2145
07:50:09-552269 DEBUG FastAPI middleware: ['Middleware', 'Middleware']
07:50:09-569112 DEBUG Creating API
07:50:09-791200 INFO [AgentScheduler] Task queue is empty
07:50:09-794668 INFO [AgentScheduler] Registering APIs
07:50:09-951173 DEBUG Scripts setup: ['IP Adapters:0.022', 'AnimateDiff:0.014', 'X/Y/Z Grid:0.015', 'Face:0.024', 'Image-to-Video:0.007', 'Stable Video Diffusion:0.006']
07:50:09-954564 DEBUG Model metadata: file="metadata.json" no changes
07:50:09-958802 DEBUG Model requested: fn=<lambda>
07:50:09-961732 WARNING Cannot generate without a checkpoint
07:50:09-964534 INFO Set system paths to use existing folders in a different location
07:50:09-967381 INFO or use --models-dir <path-to-folder> to specify base folder with all models
07:50:09-970234 INFO or use --ckpt-dir <path-to-folder> to specify folder with sd models
07:50:09-973054 INFO or use --ckpt <path-to-checkpoint> to force using existing model
07:50:09-976588 DEBUG Script callback init time: image_browser.py:ui_tabs=0.59 system-info.py:app_started=0.08 task_scheduler.py:app_started=0.18
07:50:09-980030 INFO Startup time: 37.11 torch=12.96 olive=0.21 gradio=2.65 libraries=4.92 ldm=0.14 samplers=0.13 extensions=8.56 face-restore=2.86 upscalers=0.43 networks=0.12 ui-en=0.36 ui-txt2img=0.08 ui-img2img=0.08
ui-control=0.15 ui-extras=0.22 ui-models=0.08 ui-settings=0.64 ui-extensions=1.33 ui-defaults=0.09 launch=0.58 api=0.13 app-started=0.26
07:50:09-987072 DEBUG Save: file="config.json" json=18 bytes=721 time=0.011 This time I deleted the entire Conda environment and compiled torch, torchvision and installed SDNext. I also checked there is torch installed outside conda environment. Inside Conda
(SD) sinfinities@sys5:~/anaconda3/lib$ python -c "import torch; print(torch.__version__)"
2.0.0a0+gite9ebda2Outside Conda
(base) sinfinities@sys5:~$ python -c "import torch; print(torch.__version__)"
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/home/sinfinities/Downloads/pytorch/torch/__init__.py", line 443, in <module>
raise ImportError(textwrap.dedent('''
ImportError: Failed to load PyTorch C extensions:
It appears that PyTorch has loaded the `torch/_C` folder
of the PyTorch repository rather than the C extensions which
are expected in the `torch._C` namespace. This can occur when
using the `install` workflow. e.g.
$ python setup.py install && python -c "import torch"
This error can generally be solved using the `develop` workflow
$ python setup.py develop && python -c "import torch" # This should succeed
or by running Python from a different directory.Or am not noticing something which I should?
Disty0
commented
6 months ago SDNext won't use Conda. It has it's own venv. You can remove Conda if you want.
Activate the venv of SDNext, then install PyTorch.
source venv/bin/activateIs there any way to let SDNext to use Conda? Because the dependencies which Pytorch 2.0 requires bit older and conda has the libraries of them such as C and C++ 9.5, Cuda toolkit11.4 etc. Debian 12 does not have all of them. I might be asking a rookie question, but I do not know and that's why I am asking.
vladmandic
commented
6 months ago there are three parts parts to startup: create/activate venv, run installer and last actual webui.
--skip-torch, torch will not get checked or installedimport torch, so whatever python chooses.venv, you can run simple python app in global context.
and if you're inside conda, that becomes your context../webui.sh use python launch.pyvladmandic
commented
6 months ago closing the issue, but if there are any additional questions post here and we can continue.
Issue Description
I do have an old Graphic card which supports cuda 11.4 I have compiled the pytroch from source to be able to use the GPU but whenever I am trying to install SDNext script running to install latest torch version, which defeats the purpose of compiling for hours. I have torch compiled in conda environment along with xformers.
If I use
python -c "import torch; print(torch.__version__); print(torch.randn(1).cuda())"I get prompt
I did try to find out from which file it's pushing for the pip install but unfortunately I have not able to find the file or the line yet. My question is how to use compiled torch instead of updated one?
Version Platform Description
No response
Relevant log output
No response
Backend
Original
Branch
Master
Model
SD 1.5
Acknowledgements