General Training Pipeline

Recommendations:

1. get the dataset curation tool
2. download newest/best MODEL, autotag model, lastest auto1111 repo, lastest koyah_ss repo
3. setup auto1111, ignore the model download at start & install the Ulitmate Upscaler extension
4. download and/or upload desired data and edit tag/s and/or image/s with the editor and save all changes
5. run auto1111, ignore the model download at start & upscale by directory all the images in the img_downloads folder in the data curation tool directory, lastly search and delete the jpgs in the destination folder; the pngs are the generated/upscaled ones
6. setup the koyah_ss trainer & go to the "Dreambooth Lora" tab, then go to the "Tools" tab, then configure repeats & other options, then run the generate data button
7. move the NEW data folder in the required directory structure needed for training
8. configure all the training options in the LORA tab & print the run command
9. copy & paste the run command into the terminal using the desired venv/conda environment (when ready)

========================= BREAKDOWN OF STEPS =========================

• Steps 1-3 acquire all repos, models, and extension/s needed for data prep & training
• Step 4 alter the dataset with the proper tag/s and image/s desired for training (there's an optional auto-tag model incase manual work is needed) with the dataset curation tool
• Step 5 alter the dataset by (un-glazing)/denoising/upscaling desired image/s with the downloaded upscaler extension in the auto11 ui
• Steps 6,7 alter the dataset with the proper repeats & folder structure needed for training
• Steps 8,9 configure training settings & train