fast-stable-diffusion
fast-stable-diffusion copied to clipboard
TheLastBen
Collab error - not working any more since today
"WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for: PyTorch 2.1.0+cu118 with CUDA 1106 (you have 2.1.0+cu121) Python 3.9.16 (you have 3.10.12) Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)"
I would really like to work without interruptions like that. What does it mean? What do I have to do? For month I worked without much problems and suddenly this. How can I fix it? I am paying for Colab. What else can I do? Really p******!!
Probably a Python update. We might add manually a update with the old version of Python and Pytorch tho (https://download.pytorch.org/whl/cu118). I tried with a new environnement, but I get an error :(
I tried the camenduru implementation - but that's another complete disaster
I tried this:
!pip install lmdb !pip install torch==2.0.1+cu118 torchvision==0.15.2+cu118 torchaudio==2.0.2 torchtext==0.15.2+cpu torchdata==0.6.1 --index-url https://download.pytorch.org/whl/cu118
was a tip from an old post from October. But it doesn't help either....
I'm giving up for tonight. Thing is: I want to pay for a valuable product. I have no unpaid apps on my computer nor my phone. I just need stuff that works. But SD is more and more like my experiences in coding in the late 80s of the last century: might work, might not. Dall-E is not flexible enough for me. Midnight Journet e.a. same. ComfyUI doesn't have a proper notebook and takes some time to get adjust. SD Next doesn't care about cloud computing (and not about MacOs anyway). So: even I would spend some $$ for an undisturbed uncluttered workflow: I CANT'T :-(
Can confirm, stopped working between when I left for work and when I came back. Same error.
In my environment, I added the following code to the beginning of the notebook and ran it, and StableDiffusion ran as before. I also made some modifications to the October Tips: [https://github.com/TheLastBen/fast-stable-diffusion/issues/2615]
!pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118
In my environment, I added the following code to the beginning of the notebook and ran it, and StableDiffusion ran as before. I also made some modifications to the October Tips: [https://github.com//issues/2615]
!pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118
You are my personal hero. But you always have to run the script again when you have finished the runtime.
@aegia13 you are officially the internet first on this one. Havent used my SD colab for a month, been googling this issue for two hours no luck. Finally realized to try here, you'd posted the fix less than 30min earlier. Total gratitude here.
sorry to say when i use highres.fix，it could not work either...
Traceback (most recent call last):
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/call_queue.py", line 36, in f
res = func(*args, **kwargs)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/txt2img.py", line 55, in txt2img
processed = processing.process_images(p)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 732, in process_images
res = process_images_inner(p)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/extensions/sd-webui-controlnet/scripts/batch_hijack.py", line 42, in processing_process_images_hijack
return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 867, in process_images_inner
samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 1156, in sample
return self.sample_hr_pass(samples, decoded_samples, seeds, subseeds, subseed_strength, prompts)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 1242, in sample_hr_pass
samples = self.sampler.sample_img2img(self, samples, noise, self.hr_c, self.hr_uc, steps=self.hr_second_pass_steps or self.steps, image_conditioning=image_conditioning)
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/sd_samplers_kdiffusion.py", line 145, in sample_img2img
sigma_sched = sigmas[steps - t_enc - 1:]
TypeError: slice indices must be integers or None or have an __index__ method
sorry to say when i use highres.fix，it could not work either...
Traceback (most recent call last): File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/call_queue.py", line 57, in f res = list(func(*args, **kwargs)) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/call_queue.py", line 36, in f res = func(*args, **kwargs) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/txt2img.py", line 55, in txt2img processed = processing.process_images(p) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 732, in process_images res = process_images_inner(p) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/extensions/sd-webui-controlnet/scripts/batch_hijack.py", line 42, in processing_process_images_hijack return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 867, in process_images_inner samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 1156, in sample return self.sample_hr_pass(samples, decoded_samples, seeds, subseeds, subseed_strength, prompts) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/processing.py", line 1242, in sample_hr_pass samples = self.sampler.sample_img2img(self, samples, noise, self.hr_c, self.hr_uc, steps=self.hr_second_pass_steps or self.steps, image_conditioning=image_conditioning) File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/modules/sd_samplers_kdiffusion.py", line 145, in sample_img2img sigma_sched = sigmas[steps - t_enc - 1:] TypeError: slice indices must be integers or None or have an __index__ method
Hires.fix work's fine for me
!pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118
sadly it doesnt work today for some reason, getting some cuda setup failed error
I added this in the requirements section, and it's working for me (it's a fix from Xela for his notebook apparently this is an error caused by colab env updates)
!pip install https://download.pytorch.org/whl/cu121/xformers-0.0.22.post4-cp310-cp310-manylinux2014_x86_64.whl#sha256=7075114dbf698b609b599f0d35032c0b2f9a389751e8bbf4dd3c628376b0dd9c
Tried all the above approaches, still getting the error, for CUDA setup.
CUDA SETUP: Setup Failed!
@Saranmt @HaikenEdge @bladewolf74 @TheLastBen anyone came up with any solution?
Not sure if this helps, but this is my current working Requirements section, tho I have to erase runtime, refresh and run from scratch all cells on each batch I run, so it's not the best fix, I'd say a temporary one:
#@markdown # Requirements
print('[1;32mInstalling requirements...')
with capture.capture_output() as cap:
%cd /content/
!wget -q -i https://raw.githubusercontent.com/TheLastBen/fast-stable-diffusion/main/Dependencies/A1111.txt
!dpkg -i *.deb
if not os.path.exists('/content/gdrive/'+mainpth+'/sd/stablediffusiond'): #restore later
!tar -C /content/gdrive/$mainpth --zstd -xf sd_mrep.tar.zst
!tar -C / --zstd -xf gcolabdeps.tar.zst
!rm *.deb | rm *.zst | rm *.txt
if not os.path.exists('gdrive/'+mainpth+'/sd/libtcmalloc/libtcmalloc_minimal.so.4'):
%env CXXFLAGS=-std=c++14
!wget -q https://github.com/gperftools/gperftools/releases/download/gperftools-2.5/gperftools-2.5.tar.gz && tar zxf gperftools-2.5.tar.gz && mv gperftools-2.5 gperftools
!wget -q https://github.com/TheLastBen/fast-stable-diffusion/raw/main/AUTOMATIC1111_files/Patch
%cd /content/gperftools
!patch -p1 < /content/Patch
!./configure --enable-minimal --enable-libunwind --enable-frame-pointers --enable-dynamic-sized-delete-support --enable-sized-delete --enable-emergency-malloc; make -j4
!mkdir -p /content/gdrive/$mainpth/sd/libtcmalloc && cp .libs/libtcmalloc*.so* /content/gdrive/$mainpth/sd/libtcmalloc
%env LD_PRELOAD=/content/gdrive/$mainpth/sd/libtcmalloc/libtcmalloc_minimal.so.4
%cd /content
!rm *.tar.gz Patch && rm -r /content/gperftools
else:
%env LD_PRELOAD=/content/gdrive/$mainpth/sd/libtcmalloc/libtcmalloc_minimal.so.4
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
os.environ['PYTHONWARNINGS'] = 'ignore'
!sed -i 's@text = _formatwarnmsg(msg)@text =\"\"@g' /usr/lib/python3.10/warnings.py
!pip install open-clip-torch==2.20.0 -qq --no-deps
!pip install fastapi==0.94.0 gradio==3.41.2 -qq
!pip install lmdb
!pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu121
!pip install https://download.pytorch.org/whl/cu121/xformers-0.0.22.post4-cp310-cp310-manylinux2014_x86_64.whl#sha256=7075114dbf698b609b599f0d35032c0b2f9a389751e8bbf4dd3c628376b0dd9c
clear_output()
inf('\u2714 Done','success', '50px')
#@markdown ---
Not sure if this helps, but this is my current working Requirements section, tho I have to erase runtime, refresh and run from scratch all cells on each batch I run, so it's not the best fix, I'd say a temporary one:
#@markdown # Requirements print('�[1;32mInstalling requirements...') with capture.capture_output() as cap: %cd /content/ !wget -q -i https://raw.githubusercontent.com/TheLastBen/fast-stable-diffusion/main/Dependencies/A1111.txt !dpkg -i *.deb if not os.path.exists('/content/gdrive/'+mainpth+'/sd/stablediffusiond'): #restore later !tar -C /content/gdrive/$mainpth --zstd -xf sd_mrep.tar.zst !tar -C / --zstd -xf gcolabdeps.tar.zst !rm *.deb | rm *.zst | rm *.txt if not os.path.exists('gdrive/'+mainpth+'/sd/libtcmalloc/libtcmalloc_minimal.so.4'): %env CXXFLAGS=-std=c++14 !wget -q https://github.com/gperftools/gperftools/releases/download/gperftools-2.5/gperftools-2.5.tar.gz && tar zxf gperftools-2.5.tar.gz && mv gperftools-2.5 gperftools !wget -q https://github.com/TheLastBen/fast-stable-diffusion/raw/main/AUTOMATIC1111_files/Patch %cd /content/gperftools !patch -p1 < /content/Patch !./configure --enable-minimal --enable-libunwind --enable-frame-pointers --enable-dynamic-sized-delete-support --enable-sized-delete --enable-emergency-malloc; make -j4 !mkdir -p /content/gdrive/$mainpth/sd/libtcmalloc && cp .libs/libtcmalloc*.so* /content/gdrive/$mainpth/sd/libtcmalloc %env LD_PRELOAD=/content/gdrive/$mainpth/sd/libtcmalloc/libtcmalloc_minimal.so.4 %cd /content !rm *.tar.gz Patch && rm -r /content/gperftools else: %env LD_PRELOAD=/content/gdrive/$mainpth/sd/libtcmalloc/libtcmalloc_minimal.so.4 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' os.environ['PYTHONWARNINGS'] = 'ignore' !sed -i 's@text = _formatwarnmsg(msg)@text =\"\"@g' /usr/lib/python3.10/warnings.py !pip install open-clip-torch==2.20.0 -qq --no-deps !pip install fastapi==0.94.0 gradio==3.41.2 -qq !pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu121 !pip install https://download.pytorch.org/whl/cu121/xformers-0.0.22.post4-cp310-cp310-manylinux2014_x86_64.whl#sha256=7075114dbf698b609b599f0d35032c0b2f9a389751e8bbf4dd3c628376b0dd9c clear_output() inf('\u2714 Done','success', '50px') #@markdown ---
There is a variable mainpth that you are using. what is it?
What are the libraries or import command that you added can you highlight that.
Also what do you mean by temporary fix?
Not sure if this helps, but this is my current working Requirements section, tho I have to erase runtime, refresh and run from scratch all cells on each batch I run, so it's not the best fix, I'd say a temporary one:
#@markdown # Requirements print('�[1;32mInstalling requirements...') with capture.capture_output() as cap: %cd /content/ !wget -q -i https://raw.githubusercontent.com/TheLastBen/fast-stable-diffusion/main/Dependencies/A1111.txt !dpkg -i *.deb if not os.path.exists('/content/gdrive/'+mainpth+'/sd/stablediffusiond'): #restore later !tar -C /content/gdrive/$mainpth --zstd -xf sd_mrep.tar.zst !tar -C / --zstd -xf gcolabdeps.tar.zst !rm *.deb | rm *.zst | rm *.txt if not os.path.exists('gdrive/'+mainpth+'/sd/libtcmalloc/libtcmalloc_minimal.so.4'): %env CXXFLAGS=-std=c++14 !wget -q https://github.com/gperftools/gperftools/releases/download/gperftools-2.5/gperftools-2.5.tar.gz && tar zxf gperftools-2.5.tar.gz && mv gperftools-2.5 gperftools !wget -q https://github.com/TheLastBen/fast-stable-diffusion/raw/main/AUTOMATIC1111_files/Patch %cd /content/gperftools !patch -p1 < /content/Patch !./configure --enable-minimal --enable-libunwind --enable-frame-pointers --enable-dynamic-sized-delete-support --enable-sized-delete --enable-emergency-malloc; make -j4 !mkdir -p /content/gdrive/$mainpth/sd/libtcmalloc && cp .libs/libtcmalloc*.so* /content/gdrive/$mainpth/sd/libtcmalloc %env LD_PRELOAD=/content/gdrive/$mainpth/sd/libtcmalloc/libtcmalloc_minimal.so.4 %cd /content !rm *.tar.gz Patch && rm -r /content/gperftools else: %env LD_PRELOAD=/content/gdrive/$mainpth/sd/libtcmalloc/libtcmalloc_minimal.so.4 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' os.environ['PYTHONWARNINGS'] = 'ignore' !sed -i 's@text = _formatwarnmsg(msg)@text =\"\"@g' /usr/lib/python3.10/warnings.py !pip install open-clip-torch==2.20.0 -qq --no-deps !pip install fastapi==0.94.0 gradio==3.41.2 -qq !pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu121 !pip install https://download.pytorch.org/whl/cu121/xformers-0.0.22.post4-cp310-cp310-manylinux2014_x86_64.whl#sha256=7075114dbf698b609b599f0d35032c0b2f9a389751e8bbf4dd3c628376b0dd9c clear_output() inf('\u2714 Done','success', '50px') #@markdown ---There is a variable mainpth that you are using. what is it?
What are the libraries or import command that you added can you highlight that.
Also what do you mean by temporary fix?
the main path is in the original Automatic1111 I didnt change anything and have no idea what it does, the only part I added was this: !pip install https://download.pytorch.org/whl/cu121/xformers-0.0.22.post4-cp310-cp310-manylinux2014_x86_64.whl#sha256=7075114dbf698b609b599f0d35032c0b2f9a389751e8bbf4dd3c628376b0dd9c
,,, by temporary fix , I mean this is not a good fix since I have to rerun it from scratch everytime, hopefully someone comes up with a better solution
In my environment, I added the following code to the beginning of the notebook and ran it, and StableDiffusion ran as before. I also made some modifications to the October Tips: [https://github.com//issues/2615]
!pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118
after testing a while I have to say: It worked for me. Thanks aegia13, saves my day :-)
ng a while I @MajorGruberth Can you share me how it worked for you? Also if you could share the colab file it would be great.
In my environment, I added the following code to the beginning of the notebook and ran it, and StableDiffusion ran as before. I also made some modifications to the October Tips: [https://github.com//issues/2615] !pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118
after testing a while I have to say: It worked for me. Thanks aegia13, saves my day :-)
just to clarify were you using fast-dreambooth or fast-stable_diffusion it doesnt seem to work with the former but I just want to confirm
!pip uninstall xformers
!pip uninstall torch
!pip install xformers
!pip install torch
solves it for me in other projects as well
https://colab.research.google.com/drive/17A41komGbvrhY-0mNFd-ZuNQKNONXGdp?usp=sharing
What is worse is that is a general error. The colab version of Kohya trainer also stopped working.
What is worse is that is a general error. The colab version of Kohya trainer also stopped working.
Yeah it maybe because of the colab version upgrade or some package library upgrade. But need solution if anyone comes up with.
ng a while I @MajorGruberth Can you share me how it worked for you? Also if you could share the colab file it would be great.
Certainly. I modified the first part of the notebook, "Connect Google Drive" and it looks like:
#@markdown # Connect Google Drive !pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118 from google.colab import drive from IPython.display import clear_output import ipywidgets as widgets import os
def inf(msg, style, wdth): inf = widgets.Button(description=msg, disabled=True, button_style=style, layout=widgets.Layout(min_width=wdth));display(inf) Shared_Drive = "" #@param {type:"string"} #@markdown - Leave empty if you're not using a shared drive
print("[0;33mConnecting...") drive.mount('/content/gdrive')
if Shared_Drive!="" and os.path.exists("/content/gdrive/Shareddrives"): mainpth="Shareddrives/"+Shared_Drive else: mainpth="MyDrive"
clear_output() inf('\u2714 Done','success', '50px')
#@markdown ---
with other words you click "show code" and paste in:
!pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118
than start the notebook
What is worse is that is a general error. The colab version of Kohya trainer also stopped working.
https://github.com/TheLastBen/fast-stable-diffusion/issues/2681#issuecomment-1857881050 , should be a (temporary) general solution
What is worse is that is a general error. The colab version of Kohya trainer also stopped working.
#2681 (comment) , should be a (temporary) general solution
Just this did not worked.
It showed Exception: CUDA SETUP: Setup Failed!
I just added 'xformers' also at the end and now it works (wasn't the case without it):
#@markdown # Connect Google Drive !pip install lmdb !pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 torchaudio==2.1.0 torchtext==0.16.0+cpu torchdata==0.7.0 --index-url https://download.pytorch.org/whl/cu118 xformers
