Credit to Gustavosta on Hugging Face for model

Based on Gustavosta's app.py script used for his Hugging Face Spaces. This is a trimmed and simplified implementation intended for running locally directly with python or via notebook

Requirements:

- Python 3.4 or higher
- Git-LFS 
- (Linux): python pip, tkinter
	- Tested on:
	- arch based: pacman -S python-pip tk
	- ubuntu 22.04 based: apt install python3-pip python3-tk

How to use:

Scripts

- (Windows) Seperate batch files for python, jupyter, and colab local runtime
- (Linux) Single batch script with launch arguments
	- [no arguments] installs base requirements and launches normally
	- [ --jupyter ] installs base/jupyter requirements, launches notebook in browser
	- [ --colab ] installs base/jupyter requirements, allows local runtime from Google Colab

Python

- From terminal use " pip install -r requirements.txt " to grab dependencies
- Run python MagicPrompt.py
- Visit localhost:8088 in browser (substitute 'localhost' with devices ip address if run remotely)

Jupyter Notebook

- From terminal use " pip install -r notebook-requirements.txt " to grab dependencies
- Run " jupyter notebook MagicPrompt.ipynb "

Google Colab

- Visit https://colab.research.google.com
- Choose Upload tab
- Browse to file " MagicPrompt.ipynb "

- ***  Using Google's free servers: ***
- Just click connect in top right
- Use Runtime menu to run all

- ***  For local runtime:  ***
- From terminal use " pip install -r notebook-requirements.txt " to grab dependencies
- Run " jupyter notebook --no-browser --NotebookApp.allow_origin="https://colab.research.google.com" --port=8888 --NotebookApp.port_retries=0 "
- Copy URL containing token (e.g. http://localhost:8888/?token=5bbec....)
- In Colab, use dropdown next to connect and select 'Connect to a local runtime'
- Paste URL into the 'Backend URL' box and click connect
- Use Runtime menu to run all

From original README.md

This is a model from the MagicPrompt series of models, which are GPT-2 models intended to generate prompt texts for imaging AIs, in this case: Stable Diffusion. This model was trained with 150,000 steps and a set of about 80,000 data filtered and extracted from the image finder for Stable Diffusion: "Lexica.art". It was a little difficult to extract the data, since the search engine still doesn't have a public API without being protected by cloudflare, but if you want to take a look at the original dataset, you can have a look here: datasets/Gustavosta/Stable-Diffusion-Prompts. If you want to test the model with a demo, you can go to: "spaces/Gustavosta/MagicPrompt-Stable-Diffusion".

Changes from 'Spaces' app.py

Model is saved to local folder allowing for offline use
Colab-style notebook file for easy editing and use of gradio's in line interface
Enabled gradio flagging feature, allows saving of prompt to log.csv
Trimmed code by switching gradio's '.interface()' to '.interface.from_pipeline'
- This was done to simplify code and prevent log.csv from filling with unwanted prompts
- Likely going to change once I better learn python and the options of the transformers library
Added requirements.txt for quick terminal deployment via pip
Also added a seperate notebook-requirements.txt for use with jupyter notebook
Changed port from gradio's standard to allow running at the same time as another gradio interface (e.g. Automatic1111's WebUI using default ports)
Changed listening address from '127.0.0.1' (local host only) to '0.0.0.0' (network wide)
- allows for access from other devices, useful for deployment on headless server
Removed use of ideas.txt for examples
- While some would consider it a nice addition, I find no value for my use case
Added a favicon 🤗

Planned

Add gradio options to test transformers temperature setting
Add gradio options to test transformers beam search
Add gradio options to select number of return sequences
Add gradio option for random seed selection
Add gradio tab for log.csv entries
Consolidate .bat files into single file with launch args
Add launch args for port selection

Pull requests welcome! I have no former python, and minimal scripting, experience.

This was intended as a learning project to get into python.

💻 You can see other MagicPrompt models:

For Dall-E 2: Gustavosta/MagicPrompt-Dalle
For Midjourney: Gustavosta/MagicPrompt-Midourney [⚠️ In progress]
MagicPrompt full: Gustavosta/MagicPrompt [⚠️ In progress]

⚖️ Licence:

MIT

When using this model, please credit: Gustavosta

Thanks for reading this far! :)

RassilonSleeps/MagicPrompt-SD