Comments (7)
the problem with auto1111 is torch version 1.13.1+cu117. dreambooth only works currently with torch version 1.13.1+cu116
as a gpu solution on the cheap try to find a p40 on ebay. its the pascal titan version with 24gb for enterprise as a compute card and can be had for 200ish dollars on ebay
from kohya_ss.
I wish I had the time... But I don't use colab and would not have the passion to maintain it. If someone create a version I will gladly point to it in the readme.
from kohya_ss.
I wish I could, but I hate Python with a passion, and it would require a colab with a cli.
I tried using the notebook up there with LoRA and 5 hours later everything was an utter failure YET someone I know did fantastic with this on his 3090 YET using the same settings with the notebook I wasted all that time to have nothing to show for it.
from kohya_ss.
I understand. I used to run colab and hated every minutes of it... Hence why I invested in a 3090 so I could free myself from remote GPUs.
from kohya_ss.
I understand. I used to run colab and hated every minutes of it... Hence why I invested in a 3090 so I could free myself from remote GPUs.
They insulate us far too much, and I feel like I am in a box when I use them, plus every minute I feel extremely rushed because my 2.5h per day is ticking away. Another thing is to help drain the time away I have to constantly redownload, and update, everything (seems all notebooks do this) so by the time it actually starts I wasted 10 minutes right there.
Considering what Nvidia has done with their utter greed my saving for what was supposed to be the 16GB 4060 for MSRP $379 was destroyed so when that happened I went to colab. With everything that has happened, and all the notebooks being mostly, to completely, broken, I have almost lost all interest in this stuff now.
from kohya_ss.
Alrighty, then how about an Automatic1111 webgui extension? If it is an extension I can run that on Colab (everyone can), and you need not worry about colab at all (as all my other extensions)? The Dreambooth LoRA trainer is not nearly as good AND that dev went nuts and started closing tickets, and telling his users off when they submitted a ticket as well as letting the world know some bad language stuff. I bet it could be done considering both are webgui gradio based.
from kohya_ss.
i hope you can do this i try to do but i have many errors but the base maybe work for do this implementation i share the code
!git clone https://github.com/bmaltais/kohya_ss.git
%cd kohya_ss
!pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116
!pip install --use-pep517 --upgrade -r requirements.txt
!pip install -U -I --no-deps https://github.com/C43H66N12O12S2/stable-diffusion-webui/releases/download/f/xformers-0.0.14.dev0-cp310-cp310-win_amd64.whl
!wget -N https://github.com/korakot/kora/releases/download/v0.10/py310.sh
!bash ./py310.sh -b -f -p /usr/local
!python -m ipykernel install --name "py310" --user
!conda install -c "xformers/label/dev" xformers -y
!pip install open-clip-torch
!pip install triton
!pip install tensorflow-gpu==2.3.0
!pip install tensorflow-compression
!pip install tensorrt
!pip install --use-pep517 --upgrade -r requirements.txt
%cd /content/sample_data/kohya_ss/venv/lib/python3.10/site-packages
!pip install accelerate
!conda install -c conda-forge accelerate
!pip install git+https://github.com/huggingface/accelerate
!git clone https://github.com/huggingface/accelerate
%cd accelerate
!pip install -e .
#!accelerate config
!python -c "from accelerate.utils import write_basic_config; write_basic_config(mixed_precision='fp16')"
!accelerate env
import gradio as gr
import os
import argparse
from dreambooth_gui import dreambooth_tab
from finetune_gui import finetune_tab
from textual_inversion_gui import ti_tab
from library.utilities import utilities_tab
from library.extract_lora_gui import gradio_extract_lora_tab
from library.merge_lora_gui import gradio_merge_lora_tab
from lora_gui import lora_tab
def UI(username, password):
css = ''
if os.path.exists('./style.css'):
with open(os.path.join('./style.css'), 'r', encoding='utf8') as file:
print('Load CSS...')
css += file.read() + '\n'
interface = gr.Blocks(css=css, title="Kohya_ss GUI")
with interface:
with gr.Tab('Dreambooth'):
(
train_data_dir_input,
reg_data_dir_input,
output_dir_input,
logging_dir_input,
) = dreambooth_tab()
with gr.Tab('Dreambooth LoRA'):
lora_tab()
with gr.Tab('Dreambooth TI'):
ti_tab()
with gr.Tab('Finetune'):
finetune_tab()
with gr.Tab('Utilities'):
utilities_tab(
train_data_dir_input=train_data_dir_input,
reg_data_dir_input=reg_data_dir_input,
output_dir_input=output_dir_input,
logging_dir_input=logging_dir_input,
enable_copy_info_button=True,
)
gradio_extract_lora_tab()
gradio_merge_lora_tab()
# Show the interface
if not username == '':
interface.launch(auth=(username, password))
else:
interface.launch(debug=True, share=True)
if name == 'main':
# torch.cuda.set_per_process_memory_fraction(0.48)
UI(username='', password='')
parser.add_argument(
'--username', type=str, default='', help='Username for authentication'
)
parser.add_argument(
'--password', type=str, default='', help='Password for authentication'
)
args = parser.parse_args()
UI(username=args.username, password=args.password)
from kohya_ss.
Related Issues (20)
- My model files come in json instead of a safetensor HOT 8
- Cant train SDXL HOT 1
- new error on runpod HOT 5
- Default Max Train steps is 1600 and not zero - in v24.0,7 HOT 2
- Error when starting LoRa training: FileNotFoundError: [WinError 3] The system cannot find the path specified: '' HOT 1
- LR Scheduler adafactor error / help needed (Kohya_ss GUI template on Runpod/ashleykza/kohya:24.0.6) HOT 2
- Can't train a lora for SDXL from a fresh install. HOT 5
- ValueError: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost. HOT 4
- How can I train a lora without instance prompt and class token? HOT 2
- --always_first_tags is not passwed when using WD14 Captioning HOT 7
- lion optimiser extra args
- No module pkg_resources HOT 1
- [Feature Request] Print Command - Save TOML file as named file instead of temp file HOT 8
- [Feature Request] Support for simultaneous loading of tag and caption files. HOT 2
- Cannot copy cli accelerate command and args from Print--please bring back command in Print HOT 2
- Resume training from LyCORIS HOT 1
- Can't set custom caption file extention HOT 2
- ValueError: no metadata / メタデータファイルがありません: /meta_cap.json HOT 3
- PermissionError: [Errno 13] Permission denied default_config.yaml HOT 1
- Error when creating a lora HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from kohya_ss.