Software Engineering Courses (SECourses)•3y ago

yes i did that part as well! wow thanks for replying!!! the only part that I can see that is differe

yes i did that part as well! wow thanks for replying!!! the only part that I can see that is different from your tutorial is that when creating the pod, the template i could use was this one (as opposed to the 8.0.3 version in your (excellent:1.2):) tutorial)

FFurkan Gözükara SECourses hello. did you use auto installer?

Brofessor•7/12/23, 1:13 PM

i used manual commands, but they were same as in autoinstaller

Brofessor•7/12/23, 1:13 PM

apt update
yes | apt-get install python3.10-tk

cd /workspace

git clone https://github.com/bmaltais/kohya_ss.git

cd /workspace/kohya_ss

#git checkout 429d2f282fb32e7c63b5e5fe6dbc1ddef92bba58

python3 -m venv venv

source venv/bin/activate

yes | apt-get install python3.10-tk

./setup.sh -n

GitHub

GitHub - bmaltais/kohya_ss

Contribute to bmaltais/kohya_ss development by creating an account on GitHub.

Brofessor•7/12/23, 1:14 PM

i think this tensotrt error has been around before, i checked the chat

Rrevoerom yes i did that part as well! wow thanks for replying!!! the only part that I can...

Furkan Gözükara SECourses•7/12/23, 1:16 PM

Ye they updated

Rrevoerom yes i did that part as well! wow thanks for replying!!! the only part that I can...

Furkan Gözükara SECourses•7/12/23, 1:17 PM

Let me verify

BBrofessor i used manual commands, but they were same as in autoinstaller

Furkan Gözükara SECourses•7/12/23, 1:17 PM

You can do this which I did

Furkan Gözükara SECourses•7/12/23, 1:17 PM

Print command

Furkan Gözükara SECourses•7/12/23, 1:17 PM

Actually which training you doing?

Furkan Gözükara SECourses•7/12/23, 1:18 PM

Lora dreambooth or textual inversion

Brofessor•7/12/23, 1:18 PM

dreamooth

Brofessor•7/12/23, 1:18 PM

13:16:43-292053 WARNING Package wrong version: accelerate 0.19.0 required 0.15.0
13:16:43-297004 INFO Installing package: accelerate==0.15.0 tensorboard==2.12.1 tensorflow==2.12.0 wheel
13:16:56-445976 INFO Installing modules from requirements.txt...
13:16:56-493026 INFO Installing package: -e .
13:17:34-235903 INFO Configuring accelerate...
13:17:34-243414 WARNING Could not automatically configure accelerate. Please manually configure accelerate with the
option in the menu or with: accelerate config.
Copying accelerate config file to: /root/.cache/huggingface/accelerate/default_config.yaml

Brofessor•7/12/23, 1:18 PM

there is some accelerate version error

BBrofessor 13:16:43-292053 WARNING Package wrong version: accelerate 0.19.0 required 0.15....

Furkan Gözükara SECourses•7/12/23, 1:18 PM

OK your kohya not up to date

Brofessor•7/12/23, 1:18 PM

I am actually trying to reinstall everything from scracth

Furkan Gözükara SECourses•7/12/23, 1:18 PM

You need reinstall

Brofessor•7/12/23, 1:18 PM

git clone?

Furkan Gözükara SECourses•7/12/23, 1:18 PM

Do this

Furkan Gözükara SECourses•7/12/23, 1:19 PM

Rm -r kohya_ss

Furkan Gözükara SECourses•7/12/23, 1:19 PM

Run auto installer again

Furkan Gözükara SECourses•7/12/23, 1:19 PM

It will install from scratch. It is working I am using today

Furkan Gözükara SECourses•7/12/23, 1:19 PM

Doing sdxl training

Brofessor•7/12/23, 1:20 PM

nice

Brofessor•7/12/23, 1:20 PM

ive seen some very cool images with sdxl

Furkan Gözükara SECourses•7/12/23, 1:20 PM

Ye trying to find some good eetgins

Furkan Gözükara SECourses•7/12/23, 1:20 PM

Settings

Brofessor•7/12/23, 1:21 PM

i am trying to learn kohya for loras actually

Brofessor•7/12/23, 1:21 PM

because the dreambooth traning that i am using produces a lot of mutated results for certain images

Brofessor•7/12/23, 1:21 PM

like pepe

Brofessor•7/12/23, 1:21 PM

these are ugly jpegs, so probably SD models are not good with reproducing these

Brofessor•7/12/23, 1:22 PM

and models take a lot of space, otherwise the dreambooth works fine

Brofessor•7/12/23, 1:23 PM

like this

VeryNew•7/12/23, 1:37 PM

I had "reset" my webui. Some extensions go out with the problems.

I am searching the extension that put 4 little icons in the image from extra network.
I don't remember his name to install.

Thanks all !

BBrofessor i am trying to learn kohya for loras actually

Furkan Gözükara SECourses•7/12/23, 1:43 PM

nice

VVeryNew I had "reset" my webui. Some extensions go out with the problems. I am searchin...

Furkan Gözükara SECourses•7/12/23, 1:43 PM

hmm i couldnt remember this extension

Pawix•7/12/23, 1:51 PM

My checkpoints aren't saving automatically when training dreambooth, I used Ashleyk's runpod template this time.
To save a checkpoint I have to cancel training and click "Generate ckpt" button.
What could be the reason of that?
I attached my settings, "save model frequency" is set to 2. The settings are correctly saved, I loaded them before taking the screenshot.

If it can't be solved easily I'll probably switch back to https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/How-To-Install-DreamBooth-Extension-On-RunPod.md

FFurkan Gözükara SECourses nice

Brofessor•7/12/23, 1:52 PM

i think the problem was that I used a non standard model, so it could not load it. I used protagen, last version.safetensors

Brofessor•7/12/23, 1:52 PM

with standard one it worked

PPawix My checkpoints aren't saving automatically when training dreambooth, I used Ashl...

Furkan Gözükara SECourses•7/12/23, 1:59 PM

use my auto installer working great

BBrofessor i think the problem was that I used a non standard model, so it could not load i...

Furkan Gözükara SECourses•7/12/23, 1:59 PM

i see

Furkan Gözükara SECourses•7/12/23, 1:59 PM

sdxl training test

Furkan Gözükara SECourses•7/12/23, 1:59 PM

weilu•7/12/23, 2:03 PM

pipe.load_lora_weights(".", model_path)
File "/root/autodl-tmp/tools/demo/venv/lib/python3.10/site-packages/diffusers/configuration_utils.py", line 137, in getattr
raise AttributeError(f"'{type(self).name}' object has no attribute '{name}'")
AttributeError: 'StableDiffusionXLPipeline' object has no attribute 'load_lora_weights'

weilu•7/12/23, 2:04 PM

pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-0.9", torch_dtype=torch.float16)

model_path = "./models/last-000010.safetensors"
pipe.load_lora_weights(".", model_path)