Great stuff... I couldnt get my face to work on Lora on first test but I will try again with the lon
Great stuff... I couldnt get my face to work on Lora on first test but I will try again with the longer kohya video
Fi think audio book is really good
Fi am listening and it is relaxing
Fi will also add all languages subtiles
Ftorch is tied to cuda, not to automatic
Mis it OK to ask a question about the Lora training video here?
KHi Dr. Furkan and everyone, how did you solve the problem of XYZ producing the same images for all models? When I type "woman" the results look the same, even the first epoch version
KI turned off the XYZ plot and tried using the same seed for each model. They create different images.
But when I use XYZ (with or without the same seed) it still show same images for all models
But when I use XYZ (with or without the same seed) it still show same images for all models
KFor some reasons the XYZ got stuck on 1 model
AFor anyone considering using the RunPod endpoints for training, I recommend not doing so. They are very expensive and also not very reliable. You pay $0.001 per second for Dreambooth training, then the job gets stuck for 45 minutes without even doing anything and you get charged almost $3 for nothing. Running pods is much cheaper, much more stable, and much more reliable, and you are in full control of everything.
Mfollowing this video tutorial
Zero To Hero Stable Diffusion DreamBooth Tutorial By Using Automatic1111 Web UI - Ultra Detailed
but when I clicked on "Load from" it is showing an error.
Does, anyone knows how to tackle this issue
Zero To Hero Stable Diffusion DreamBooth Tutorial By Using Automatic1111 Web UI - Ultra Detailed
but when I clicked on "Load from" it is showing an error.
Does, anyone knows how to tackle this issue
Ahello i get this error when i try to start kohya ss on runpod root@b12c8b7fa1ad:/workspace/kohya_ss# bash gui.sh --share
Traceback (most recent call last):
File "/workspace/kohya_ss/tools/validate_requirements.py", line 5, in <module>
from packaging.requirements import Requirement
ModuleNotFoundError: No module named 'packaging'
root@b12c8b7fa1ad:/workspace/kohya_ss#
Traceback (most recent call last):
File "/workspace/kohya_ss/tools/validate_requirements.py", line 5, in <module>
from packaging.requirements import Requirement
ModuleNotFoundError: No module named 'packaging'
root@b12c8b7fa1ad:/workspace/kohya_ss#
Fdev branch of automatic1111. they made a new release so do a git pull
Fyep runpod is still best option with good gpu and good prices. i wouldnt trust their endpoint though. do setup with my video it works :d
FYouTubeSECourses
Discord : https://bit.ly/SECoursesDiscord. Torch 2 / Pytorch 2 is now supported along with new DreamBooth Automatic1111 Web UI extension. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 
https://www.patreon.com/SECourses
In this video, I am explaining how you can man...
https://www.patreon.com/SECoursesIn this video, I am explaining how you can man...
Flooks like you didnt follow venv part correctly. it should activate venv
Ayes i think i know what caused it, didn't run in the end
./setup.sh -n
./setup.sh -n
Awill do it now and see how it goes
Mjust watching it currently, git cloning is finished from my end
AYep, just running a few tests to their dreambooth endpoint yesterday cost me $17, dedicated GPU pods can run for a little more than 2 days non stop for that price.
Aworks
Fwow 17$ huge
Fyou could get like 50+ hours
Ayes exactly 
Aso i tried to train, it didn't work, i think it was something either with torch or xformers, did a pip uninstall/install xformers, now i try to start kohya and get this root@b12c8b7fa1ad:/workspace/kohya_ss# bash gui.sh --share
Validating that requirements are satisfied.
Error: The following packages have the wrong version:
- timm==0.6.12 (expected version torch==1.12.1, found version 2.0.1)
Run upgrade.sh or pip install -U -r /workspace/kohya_ss/requirements.txt to resolve the missing requirements listed above...
root@b12c8b7fa1ad:/workspace/kohya_ss#
Validating that requirements are satisfied.
Error: The following packages have the wrong version:
- timm==0.6.12 (expected version torch==1.12.1, found version 2.0.1)
Run upgrade.sh or pip install -U -r /workspace/kohya_ss/requirements.txt to resolve the missing requirements listed above...
root@b12c8b7fa1ad:/workspace/kohya_ss#
F@AlaBeat did you make venv in that folder? like in video
Ayes i have made venv
Fbut it is using your system wide torch version
Fnot using your venv
Fif you are our patreon supporter i can connect your pc and install
Ai will install again torch in venv folder, even though i've done it already
Fsomething missing. it should use venv
Fnot system wide default venv
AMake sure you activate your venv, just creating it does not activate it automatically
Awill do it again, got an error saying i'm out of space
Fye you need to increase runtime space because temp files will take space
Aok....
message.txt6KB
Ftry a fresh pod i think. /workspace/kohya_ss/venv/bin/python3: No module named pip
Aok now how to fix this error when i start the train
message.txt4.35KB
Fhow much vram you have?
Fah you are on runpod
Fdont use xformers
Ai tried once with xformers installed and it worked, on runpod... will try a new pod without installing xformers
Ado i do pip uninstall xformers on the new runpod too?
Awhen using this current runpod, not the new one, i get this thing (not error) WARNING:xformers:WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
PyTorch 2.0.1+cu118 with CUDA 1108 (you have 1.12.1+cu116)
when not using xformers i get oom errors with 3090...
PyTorch 2.0.1+cu118 with CUDA 1108 (you have 1.12.1+cu116)
when not using xformers i get oom errors with 3090...
Athat was why i want to use xformers
Aoh well.. fuck it no more runpod for me
A3090 has 24GB of VRAM so its an issue with your settings, you shouldn't face OOM with 24GB VRAM
AAlso fix your xformers version