Steps that I followed to make gradio_app.py work

[FBAdmirer]

Processor: INTEL 8 GEN CORE i7-8750H
Graphics Card: NVIDIA GEFORECE GTX-1070
Video MEMORY: 8GB GDDR5
System Memory: 16GB DDR4 - 2400Hz

Step 1)
conda create --prefix X:\ominicontrolgpenv python=3.10 -y (X = any drive which has more space)

Step 2)
conda activate X:\ominicontrolgpenv

Step 2)
git clone https://github.com/deepbeepmeep/OminiControlGP.git

Step 3)
cd OminiControlGP

Step 4)
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

Step 5)
Download the Triton for windows here:
https://huggingface.co/madbuda/triton-windows-builds/resolve/main/triton-3.0.0-cp310-cp310-win_amd64.whl

Step 6)
Move above downloaded triton-3.0.0-cp310-cp310-win_amd64.whl file to OminiControlGP folder

Step 7)
pip install triton-3.0.0-cp310-cp310-win_amd64.whl

Step 8)

Change the transformers in the requirements.txt file to transformers==4.49.0 in the folder OminiControlGP

Step 9)
pip install -r requirements.txt

Step 10)

from within OminiControlGP folder, type the below

python gradio_app.py --profile 5

to see the below after all the models are downloaded and after the computer is rebooted

• Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().

[mbeytekin]

Thank you. But; I followed this roadmap exactly, but I still can't get it to work.

(omini) M:\Omini>python gradio_app.py --profile 5
Traceback (most recent call last):
File "M:\Omini\gradio_app.py", line 10, in
from src.flux.generate import seed_everything, generate
File "M:\Omini\src\flux\generate.py", line 5, in
from .transformer import tranformer_forward
File "M:\Omini\src\flux\transformer.py", line 6, in
from diffusers.models.transformers.transformer_flux import (
ImportError: cannot import name 'is_torch_version' from 'diffusers.models.transformers.transformer_flux' (C:\Users\user\anaconda3\envs\omini\lib\site-packages\diffusers\models\transformers\transformer_flux.py)

[FBAdmirer]

You are welcome. I just tried again and it works for me as shown below.

(E:\ominicontrolgpenv) E:\OminiControlGP>python gradio_app.py --profile 5
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 5.26it/s]
Loading pipeline components...: 57%|█████████████████████████████▋ | 4/7 [00:01<00:01, 2.31it/s]You set add_prefix_space. The tokenizer needs to be converted from the slow tokenizers
Loading pipeline components...: 100%|████████████████████████████████████████████████████| 7/7 [00:02<00:00, 2.71it/s]
************ Memory Management for the GPU Poor (mmgp 3.1) by DeepBeepMeep ************
You have chosen the slowest profile that requires at least 24 GB of RAM and 10 GB of VRAM.
Quantization of model 'text_encoder_2' started to format 'quanto.qint8'
Quantization of model 'text_encoder_2' done
Hooked to model 'vae' (AutoencoderKL)
Hooked to model 'text_encoder' (CLIPTextTransformer)
Hooked to model 'text_encoder_2' (T5EncoderModel)
Hooked to model 'transformer' (FluxTransformer2DModel)

• Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().

---

It appears you have Linux operating system. I have Windows 11 Pro. Below are the versions for diffusers, transformers and gradio that I have. Please check yours.

(E:\OminiControlGPenv) E:\OminiControlGP>pip show diffusers
Name: diffusers
Version: 0.32.2

(E:\OminiControlGPenv) E:\OminiControlGP>pip show transformers
Name: transformers
Version: 4.49.0

(E:\OminiControlGPenv) E:\OminiControlGP>pip show gradio
Name: gradio
Version: 5.23.3