Torch version 2 actually works with Dreambooth on Linux

R

RobboRobbo•6/1/23, 7:16 AM

@Dr. Furkan Gözükara Hi, I have followed your tutorial "Easiest Way to Install & Run Stable Diffusion Web UI on PC by Using Open Source Automatic Installer" The web UI boots up, but when i try to generate an image, i get the error: safetensor_rust.SafetensorError: Error while deserializing header: MetadataIncimpleteBuffer

R

RobboRobbo•6/1/23, 7:16 AM

Can you please help?

R

Rikk•6/1/23, 9:10 AM

@Dr. Furkan Gözükara the new Adobe Firefly is truly remarkable. These are some example I made using my own photos and AI extended background.

R

Rikk•6/1/23, 9:10 AM

it even keeps the visual style if you use a concept art

R

Rikk•6/1/23, 9:10 AM

I can use it to remove the text from these Magic cards and generate the hidden part

R

Rikk•6/1/23, 9:12 AM

It's really impressive how it keeps the consistency with the rest of the image.

R

Rusch Meyer•6/1/23, 9:41 AM

Hey! Thanks for the tutorial on running Kohya on Runpod. I'm getting the ModuleNotFoundError: No module named 'tkinter' error. When I follow the further instructions i get these errors:

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
clean-fid 0.1.29 requires requests==2.25.1, but you have requests 2.28.2 which is incompatible.
open-clip-torch 2.7.0 requires protobuf==3.20.0, but you have protobuf 3.19.6 which is incompatible.
tb-nightly 2.14.0a20230511 requires google-auth-oauthlib<1.1,>=0.5, but you have google-auth-oauthlib 0.4.6 which is incompatible.
tb-nightly 2.14.0a20230511 requires tensorboard-data-server<0.8.0,>=0.7.0, but you have tensorboard-data-server 0.6.1 which is incompatible.

R

Rusch Meyer•6/1/23, 9:41 AM

ERROR: Could not find a version that satisfies the requirement torchaudio (from versions: none)
ERROR: No matching distribution found for torchaudio

R

Rusch Meyer•6/1/23, 9:42 AM

Any idea what I am doing wrong?

RRusch Meyer Hey! Thanks for the tutorial on running Kohya on Runpod. I'm getting the ModuleN...

A

ashleyOP•6/1/23, 9:52 AM

You can try:

apt-get install python3-tk

apt-get install python3-tk

apt-get install python3-tk

apt-get install python3-tk

Aashley You can try: ```apt-get install python3-tk```

R

Rusch Meyer•6/1/23, 9:58 AM

Thanks for getting back! When I run the suggested command I retrieve "python3-tk is already the newest version (3.8.10-0ubuntu1~20.04).
0 upgraded, 0 newly installed, 0 to remove and 46 not upgraded." BUT, when I start the Kohya Gui I dont get any errors any longer. When I click on the local URL i get "Unable to connect" in all browsers. Any further tips?

RRusch Meyer Thanks for getting back! When I run the suggested command I retrieve "python3-tk...

A

ashleyOP•6/1/23, 10:01 AM

It runs on port 7861 right? You will need to expose that port, you can either set your pod to have a public IP and expose the port as a TCP port, and then connect on the public IP and public port, because 7861 will be the internal port, or else you can set it as an HTTP port and then use the RunPod proxy to access it, but you will need to create a new pod if you never did that when your pod was created.

A

ashleyOP•6/1/23, 10:02 AM

You need to click this when creating your pod

A

ashleyOP•6/1/23, 10:03 AM

And then add the port either as HTTP (for proxy) or TCP (to access it from the public IP)

A

ashleyOP•6/1/23, 10:05 AM

If your pod already has port 3000 enabled, you can try something like this:

--server_port 3000

--server_port 3000

--server_port 3000

--server_port 3000

R

Rusch Meyer•6/1/23, 11:42 AM

you are my hero, thanks! First training session is running!

A

ashleyOP•6/1/23, 11:56 AM

I am thinking of creating a Khoya_ss template if anyone is interested

D

David Ce•6/1/23, 12:29 PM

Hello all, i'm new to stable diffusion and wanted to enter this intriguing subject but sadly have a low vram for dreamboothing. Are there any customizable online solutions for dreamboothing?

DDavid Ce Hello all, i'm new to stable diffusion and wanted to enter this intriguing subje...

A

ashleyOP•6/1/23, 12:38 PM

RunPod is your friend - runpod.io

A

ashleyOP•6/1/23, 12:38 PM

Use @Dr. Furkan Gözükara referral code

D

David Ce•6/1/23, 12:40 PM

hmm, alright i'll check it out thank you

D

David Ce•6/1/23, 12:40 PM

Is it better than using Hugging face spaces?

DDavid Ce Is it better than using Hugging face spaces?

A

ashleyOP•6/1/23, 12:41 PM

You have a lot more flexibility and control

D

David Ce•6/1/23, 12:43 PM

Exactly what I needed

A

ashleyOP•6/1/23, 12:43 PM

https://bit.ly/RunPodIO <--- use this link to sign up, its his referral code

Rent Cloud GPUs from $0.2/hour

Save over 80% on GPUs. GPU rental made easy with Jupyter for Tensorflow, PyTorch or any other AI framework.

A

ashleyOP•6/1/23, 12:44 PM

Lots of RunPod videos at: https://www.youtube.com/@SECourses/videos

YouTube

SECourses

Welcome to Software Engineering Courses (SECourses) – the ultimate destination for skillfully curated insights into state-of-the-art technologies and programming paradigms. We demystify the realms of Artificial Intelligence, Stable Diffusion, DreamBooth, LoRA, ControlNet, Textual Inversion, Software Engineering, Programming, C#, .NET, ASP .NET, ...

D

David Ce•6/1/23, 12:45 PM

Also is there a way to train ONE model to two image sets?. One being a face and another being an asthetic or do you have to use LoRA weights only

Aashley Lots of RunPod videos at: https://www.youtube.com/@SECourses/videos

D

David Ce•6/1/23, 12:45 PM

oh oh thank you

F

felixsanz•6/1/23, 1:04 PM

@Dr. Furkan Gözükara do you really think the model has chinesse spyware?

Ffelixsanz @Dr. Furkan Gözükara do you really think the model has chinesse spyware?

D

David Ce•6/1/23, 1:41 PM

?

F

Furkan Gözükara SECourses•6/1/23, 2:19 PM

which one better

F

Furkan Gözükara SECourses•6/1/23, 2:19 PM

F

Furkan Gözükara SECourses•6/1/23, 2:19 PM

will reply messages after publish this video

A

ashleyOP•6/1/23, 2:28 PM

New RunPod template available for Stable Diffusion Web UI:

https://runpod.io/gsc?template=ya6013lj5a&ref=w18gds2n

Includes:

Ubuntu 22.04 LTS
CUDA 11.8
Python 3.10.6
Automatic1111 Stable Diffusion Web UI 1.3.0
Dreambooth extension (dev branch)
Deforum extension
ControlNet extension v1.1.206
Torch 2.0.1
xformers 0.0.20 (disabled by default)
v1-5-pruned.safetensors
vae-ft-mse-840000-ema-pruned.safetensors

And can now also be used with Network Volumes in Secure Cloud.

A

ashleyOP•6/1/23, 2:29 PM

Torch 2.0.1 is working fine, been training Dreambooth with it all day.

A

ashleyOP•6/1/23, 2:31 PM

Just be weary of the Facebook Dadaptation optimizers. They have very specific learning rate settings compared to the others, and especially

Adan Dadaptation

Adan Dadaptation

Adan Dadaptation

Adan Dadaptation because it requires about 36GB of VRAM from my testing. The others weren't quite as intense though.

A

ashleyOP•6/1/23, 3:02 PM

Its actually very easy to set up Kohya_ss on RunPod:

apt update && apt install apt-utils
wget https://raw.githubusercontent.com/bmaltais/kohya_ss/master/setup.sh
chmod +x setup.sh
./setup.sh -d /workspace/kohya_ss
cd /workspace/kohya_ss
./gui.sh --listen 0.0.0.0 --server_port 3000

apt update && apt install apt-utils
wget https://raw.githubusercontent.com/bmaltais/kohya_ss/master/setup.sh
chmod +x setup.sh
./setup.sh -d /workspace/kohya_ss
cd /workspace/kohya_ss
./gui.sh --listen 0.0.0.0 --server_port 3000

apt update && apt install apt-utils
wget https://raw.githubusercontent.com/bmaltais/kohya_ss/master/setup.sh
chmod +x setup.sh
./setup.sh -d /workspace/kohya_ss
cd /workspace/kohya_ss
./gui.sh --listen 0.0.0.0 --server_port 3000

apt update && apt install apt-utils
wget https://raw.githubusercontent.com/bmaltais/kohya_ss/master/setup.sh
chmod +x setup.sh
./setup.sh -d /workspace/kohya_ss
cd /workspace/kohya_ss
./gui.sh --listen 0.0.0.0 --server_port 3000

It takes care of Python Tk and everything for you.

Aashley Just be weary of the Facebook Dadaptation optimizers. They have very specific l...

F

flowwolf•6/1/23, 4:34 PM

by very specific, you mean "1" set all learn rates to 1 and it takes care of it all automagically on the fly

F

flowwolf•6/1/23, 4:35 PM

i've been using dadaptlion this week and it's very succesful

F

flowwolf•6/1/23, 4:37 PM

here's a terran style i trained with it last night. not the best work. but it sure was easy work. i think improvements could be made on the dataset side but the training went so well. 20images with short minimalist captions. The train image is a good candidate for inpainting, like they're coming up to parallel rails together ready for a hijacking or something. could fixe the long train's cars too

F

flowwolf•6/1/23, 4:48 PM

N

Nom•6/1/23, 4:49 PM

@ashleyk I just spent 8 hours trying to fix dreambooth errors in runpod, and then I found this discord, I'm going to try your template to see if I can finally do some training

NNom @ashleyk I just spent 8 hours trying to fix dreambooth errors in runpod, and the...

F

flowwolf•6/1/23, 4:50 PM

not a waste of time if you consider it learning experience of what to avoid in the future

Fflowwolf not a waste of time if you consider it learning experience of what to avoid in t...

N

Nom•6/1/23, 4:50 PM

lol true. how did you do your training? also on runpod?

NNom lol true. how did you do your training? also on runpod?

F

flowwolf•6/1/23, 4:51 PM

i had a good pc already so i upgraded the gpu. big pricetag tho. used kohya's scripts, with the gui

NNom @ashleyk I just spent 8 hours trying to fix dreambooth errors in runpod, and the...

A

ashleyOP•6/1/23, 4:51 PM

Feel free to DM me if you have any questions or need help with anything.

N

Nom•6/1/23, 4:54 PM

@ashleyk how do I open the jypter page in your template, it asks for a token/password

NNom @ashleyk how do I open the jypter page in your template, it asks for a token/pas...

A

ashleyOP•6/1/23, 4:57 PM

You need to specify

JUPYTER_PASSWORD

JUPYTER_PASSWORD

JUPYTER_PASSWORD

JUPYTER_PASSWORD in the environment variables, read the README

Aashley You need to specify `JUPYTER_PASSWORD` in the environment variables, read the RE...

N

Nom•6/1/23, 5:00 PM

sorry didnt know theres a readme! I got my first error

  Traceback (most recent call last):
2023-06-01T16:59:14.685731564Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/utils/utils.py", line 200, in f
2023-06-01T16:59:14.685746164Z     res = func(*args, **kwargs)
2023-06-01T16:59:14.685748394Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/ui_functions.py", line 683, in start_training
2023-06-01T16:59:14.685750304Z     if config.pretrained_vae_name_or_path == "":
2023-06-01T16:59:14.685752134Z AttributeError: 'NoneType' object has no attribute 'pretrained_vae_name_or_path'
2023-06-01T16:59:14.685753844Z

  Traceback (most recent call last):
2023-06-01T16:59:14.685731564Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/utils/utils.py", line 200, in f
2023-06-01T16:59:14.685746164Z     res = func(*args, **kwargs)
2023-06-01T16:59:14.685748394Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/ui_functions.py", line 683, in start_training
2023-06-01T16:59:14.685750304Z     if config.pretrained_vae_name_or_path == "":
2023-06-01T16:59:14.685752134Z AttributeError: 'NoneType' object has no attribute 'pretrained_vae_name_or_path'
2023-06-01T16:59:14.685753844Z

  Traceback (most recent call last):
2023-06-01T16:59:14.685731564Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/utils/utils.py", line 200, in f
2023-06-01T16:59:14.685746164Z     res = func(*args, **kwargs)
2023-06-01T16:59:14.685748394Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/ui_functions.py", line 683, in start_training
2023-06-01T16:59:14.685750304Z     if config.pretrained_vae_name_or_path == "":
2023-06-01T16:59:14.685752134Z AttributeError: 'NoneType' object has no attribute 'pretrained_vae_name_or_path'
2023-06-01T16:59:14.685753844Z

  Traceback (most recent call last):
2023-06-01T16:59:14.685731564Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/utils/utils.py", line 200, in f
2023-06-01T16:59:14.685746164Z     res = func(*args, **kwargs)
2023-06-01T16:59:14.685748394Z   File "/workspace/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/ui_functions.py", line 683, in start_training
2023-06-01T16:59:14.685750304Z     if config.pretrained_vae_name_or_path == "":
2023-06-01T16:59:14.685752134Z AttributeError: 'NoneType' object has no attribute 'pretrained_vae_name_or_path'
2023-06-01T16:59:14.685753844Z

NNom sorry didnt know theres a readme! I got my first error 😦 ``` Traceback (most...

A

ashleyOP•6/1/23, 5:02 PM

Which model are you using as the base?

Similar Threads