stableDiffusionAutomatic1111tools

some scripts to help with Stable Diffusion, Automatic1111, Forge, and Kohya

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐

Support me

💲 My patreon: https://www.patreon.com/RobertJene

🍵 Buy me a coffee: https://www.buymeacoffee.com/robertjene

📺 Drop me a superchat on of my YouTube videos https://www.youtube.com/watch?v=oSUusZZhQPU

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐

1. embedding_inspector use this to check the strength and loss of embeddings trained in the training tab of Automatic1111 and Forge RECOMMENDED: When training an embedding, set the value "Save a copy of embedding to log directory every N steps, 0 to disable" to 50 or 100 so that the scripts have enough data to read through

   Files:

   requirements.txt

   Run the commands in this file, they are required for the script: inspect_embedding_training.py from this REPO to work: https://github.com/Zyin055/Inspect-Embedding-Training

   GPU_Stats.bat This simply utilizes GPU_Stats.sh without having to run the code in an IDE or type a command

   GPU_Stats.sh This uses nvidia-smi to query your GPU during trainging to monitor VRAM usage and GPU temperature

   embedding-show_loss+inpspect.bat Purpose: programatically uses embedding_loss.vbs (provided in this repository) and inspect_embedding_training.py which is available here: https://github.com/Zyin055/Inspect-Embedding-Training

   Menu Options: f. Set/change the log folder where training data is

   ~ use this menu to set the folder where the embedding training output is (Automatic1111 Train Tab -> Train Tab -> Log directory) NOTE: MUST be full absolute path with NO DOUBLE QUOTES (") because the script adds them itself

   1. Show Loss (find loss values below threshold)

   ~ this uses mbedding_loss.vbs, see section for that file below

   2. Create Graphs

   ~ this uses inspect_embedding_training.py to create graphs

   3. Inspect Training (strength values over 0.2 may mean overtrained)

   ~ this uses inspect_embedding_training.py to show the strength values in the embedding files

   embedding_loss.vbs This script is utilized by embedding-show_loss+inpspect.bat when they are put in the same folder. You can also use it directly with a cscript command.

   Purpose: help you filter the results of a training to show values under a certain threshold run this script after or while training an embedding

   this script accepts 1 or 2 command line arguments: 1: the path to textual_inversion_loss.csv after doing an embedding training 2: the threhold of values to show (only show values lower than this) ~ if this value is not provided, it defaults to 0.6 for the threshold

   Example commands: csript.exe "~FULL_PATH_TO_LOG_FOLDER\Model_Name_512-4950-9800" csript.exe "~FULL_PATH_TO_LOG_FOLDER\Model_Name_512-4950-9800" 0.08 csript.exe "~FULL_PATH_TO_LOG_FOLDER\Model_Name_512-4950-9800" ".04"

2. FFMPEG_Image_Converter use this when setting up training data images to convert all, or some of the files to .PNG

   Purpose: to save tons of time when converting all files to PNG format.

   there is a benefit to having all files named in this format: instancePrompt (x).png where X is a number. there is a problem when you have duplicate file names with different extensions.

   Example: instancePrompt (10).png instancePrompt (10).jpg

   this will create a problem where there is one caption file for two images files.

   Furthermore, all files should be .PNG because of the quality.

   Files:

   FFMPEG_Image_Converter.bat ~this is the option-based batch file that calls a vbscript with arguments to perform the functions.

   FFMPEG_Image_Converter_Batch_Maker.vbs ~this is the vbscript that performs the functions. ~It uses FFmpeg under the LGPLv2.1

   Menu Options: F. Change the working path. The working path is the folder with image files you are working on. D. Check for duplicate hashes. This is one method to make sure you don't have duplicates. Will only work if two image file's hases are exactly the same. You should choose this option before converting and after. E. List extensions in the folder. This option counts how many of each image file type you have in the folder. Numbered options - use when you only want to convert a single file type to PNG P. Use when you want to convert all files to PNG

   ffmpeg.exe ~not included in this repo. You can download the zip from here, and extract the file: https://ffmpeg.org make sure you put ffmpeg.exe in the same folder as the above two scripts.