c0nsume Qwen

SET UP YOUR venv:

Create a python virtual environment (tested on 3.12 but I imagine 3.9 and forward all work fine):

git clone https://github.com/E10H1M/c0nsume-qwen.git
cd c0nsume-qwen
python3 -m venv venv
. ./venv/bin/activate

SETUP FOR pytorch:

For pytorch 12.8 Linux (what i tested):

pip3 install torch torchvision

For other versions:

1. go to: https://pytorch.org/
2. scroll down
3. select the version you want. 
4. for older versions, go to: https://pytorch.org/get-started/previous-versions/

INSTALL PROJECT REQUIREMENTS:

After installing torch, install project requirements via:

pip install -r god.txt 

To download the weights for the project, refer to download_instructions.txt.
Automation for downloads will be added soon, some time tomorrow or the following day.
(Aug 20th or 21st).

RUN NOTEBOOKS:

1. Open up jupyterlab:

jupyter lab

2. Enter the /notebooks directory.

For bitsandbytes (need ~16 GB of VRAM), use

qwen_img_bnb.ipynb
qwen_edit_bnb.ipynb

For torchtao (need 24GB VRAM GPU), use

qwen_img_torchao.ipynb
qwen_edit_torchao.ipynb

---

Note:

This repo is susceptible to active changes. Especially the Readme as more model architectures are included. Next update will add Qwen-Image-Edit to the Rust/Python client server system.

Rust based client server now under /client. Refer to its README.
Python based backend server now under /server. Refer to its README.

Rest in peace Terry Davis.

Thank you for TempleOS. But even more so: for being a light.