kcg-ml-sd1p4

Summary

• kcg-ml-sd1p4
  • Prerequisites
    • Installing Requirements
    • Downloading Models
    • Processing Models
  • Running Stable Diffusion scripts
    • Text To Image
    • Embed prompts
    • Images from embeddings
    • Images from distributions
    • Images from temperature range
    • Images and encodings
    • Perturbations on prompts embeddings
    • Image Grid Generator
    • Generate Images From Prompt Generator
    • Generate Images From Prompt List Dataset
    • Chad Score
    • Chad Sort
    • Running GenerationTask
    • Prompt Score
    • Prompt Embeddings Gradient Optimization
    • Prompt Generator
    • Image Ranker by Fitness Score
    • Auto-ml
    • Generate Images From Model Predictions
    • Split images
    • Affine Combination Of Embeddings GA
    • Inpainting
    • Generate Images Using Inpainting from Prompt List Dataset
  • Tests
    • To run all scripts
    • To run all kcg-ml-sd1p4 tests

Prerequisites

Installing Requirements

Run:

pip3 install -r requirements.txt

Downloading Models

This script will download these models:

CLIP - ./input/model/clip/vit-large-patch14/model.safetensors
CLIP Text model - ./input/model/clip/text_model/pytorch_model.bin
Stable Diffusion - ./input/model/sd/v1-5-pruned-emaonly/v1-5-pruned-emaonly.safetensors

Run:

python3 ./download_models.py

Processing Models

Note: This script takes in a stable diffusion model and cuts it up into sub-models

Run:

python3 ./process_models.py

Then you already should be able to run:

python3 ./scripts/txt2img.py --num_images 2 --prompt 'A purple rainbow, filled with grass'

Running Stable Diffusion scripts

Text To Image

Takes in a prompt and generates images

These are the available CLI arguments:

options:
  -h, --help            show this help message and exit
  --prompt [PROMPT]     The prompt to render
  --negative-prompt [NEGATIVE_PROMPT]
                        The negative prompt. For things we dont want to see in generated image
  --prompts_file PROMPTS_FILE
                        Path to the file containing the prompts, each on a line (default: './input/prompts.txt')
  --batch_size BATCH_SIZE
                        How many images to generate at once (default: 1)
  --output OUTPUT       Path to the output directory (default: ./output)
  --sampler SAMPLER     Name of the sampler to use (default: ddim)
  --checkpoint_path CHECKPOINT_PATH
                        Path to the checkpoint file (default: './input/model/v1-5-pruned-emaonly.safetensors')
  --flash               whether to use flash attention
  --steps STEPS         Number of steps to use (default: 50)
  --cfg_scale CFG_SCALE
                        unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty))
  --low_vram            limit VRAM usage
  --force_cpu           force CPU usage
  --cuda_device CUDA_DEVICE
                        cuda device to use for generation
  --num_images NUM_IMAGES
                        How many images to generate (default: 1)
  --seed SEED           Seed for the image generation (default: )

Example Usage:

python3 ./scripts/text_to_image.py --prompt "character, chibi, waifu, side scrolling, white background, centered" --negative-prompt "white" --checkpoint_path "./input/model/v1-5-pruned-emaonly.safetensors" --batch_size 1 --num_images 1

Embed prompts

Saves a tensor of a batch of prompts embeddings, and the tensor for the null prompt "".

**Command line arguments**

- `-p, --prompts`: The prompts to embed. Defaults
  to `['A painting of a computer virus', 'An old photo of a computer scientist']`.
- `--embedded_prompts_dir`: The path to the directory containing the embedded prompts tensors. Defaults to a
  constant `EMBEDDED_PROMPTS_DIR`, which is expected to be `'./input/embedded_prompts/'`.

Command:

python3 ./scripts/embed_prompts.py --prompts 'A painting of a computer virus', 'An old photo of a computer scientist', 'A computer drawing a computer'

Images from embeddings

Only run this after generating the embedded prompts with the above script.

**Command line arguments**

- `-p, --embedded_prompts_dir`: The path to the directory containing the embedded prompts tensors. Defaults to
  the `EMBEDDED_PROMPTS_DIR` constant, which is expected to be `'./input/embedded_prompts/'`.
- `-od, --output_dir`: The output directory. Defaults to the `OUTPUT_DIR` constant, which is expected to
  be `'./output/noise-tests/from_embeddings'`.
- `--num_images`: Number of images to generate. Defaults to `1`.
- `-bs, --batch_size`: Batch size to use. Defaults to `1`.
- `-t, --temperature`: Sampling temperature. Defaults to `1.0`.
- `--ddim_eta`: Amount of noise to readd during the sampling process. Defaults to `0.0`.
- `--clear_output_dir`: Either to clear or not the output directory before running. Defaults to `False`.
- `--cuda_device`: CUDA device to use. Defaults to `get_device()`.
- `--low_vram`: Only use this if the gpu you have is old`.
- `--sampler`: Default value is `DDIM`.
- `--seed`: Default value is `empty string`.
- `--cfg_scale`: unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty)) . Defaults
  to `7`.
-

Try running:

python3 ./scripts/generate_images_from_embeddings.py --num_images 4 --temperature 1.2 --ddim_eta 0.2

Images from distributions

Try running:

python3 ./scripts/generate_images_from_distributions.py -d 4 --params_steps 4 --params_range 0.49 0.54 --num_seeds 4 --temperature 1.2 --ddim_eta 1.2

**Command line arguments**

- `-p, --prompt`: The prompt to generate images from. Defaults
  to `"A woman with flowers in her hair in a courtyard, in the style of Frank Frazetta"`.
- `--negative-prompt`: The negative prompt. For things we dont want to see in generated image. Defaults to `''`.
- `-od, --output_dir`: The output directory. Defaults to the `OUTPUT_DIR` constant, which should
  be `"./output/noise-tests/from_distributions"`.
- `-cp, --checkpoint_path`: The path to the checkpoint file to load from. Defaults to the `CHECKPOINT_PATH` constant,
  which should be `"./input/model/v1-5-pruned-emaonly.safetensors"`.
- `-F, --fully_initialize`: Whether to fully initialize or not. Defaults to `False`.
- `-d, --distribution_index`: The distribution index to use. Defaults to `4`. Options: 0: "Normal", 1: "Cauchy", 2: "
  Gumbel", 3: "Laplace", 4: "Logistic".
- `-bs, --batch_size`: The batch size to use. Defaults to `1`.
- `--params_steps`: The number of steps for the parameters. Defaults to `3`.
- `--params_range`: The range of parameters. Defaults to `[0.49, 0.54]`.
- `--num_seeds`: Number of random seeds to use. Defaults to `3`.
- `-t, --temperature`: Sampling temperature. Defaults to `1.0`.
- `--ddim_eta`: Amount of noise to readd during the sampling process. Defaults to `0.0`.
- `--clear_output_dir`: Either to clear or not the output directory before running. Defaults to `False`.
- `--cuda_device`: CUDA device to use. Defaults to `"get_device()"`.

Images from temperature range

Try running:

python3 ./scripts/generate_images_from_temperature_range.py -d 4 --params_range 0.49 0.54 --params_steps 3 --temperature_steps 3 --temperature_range 0.8 2.0

**Command line arguments**

- `-p, --prompt`: The prompt to generate images from. Defaults
  to `"A woman with flowers in her hair in a courtyard, in the style of Frank Frazetta"`.
- `--negative-prompt`: The negative prompt. For things we dont want to see in generated image. Defaults to `''`.
- `-od, --output_dir`: The output directory. Defaults to the `OUTPUT_DIR` constant, which is expected to
  be `"./output/noise-tests/temperature_range"`.
- `-cp, --checkpoint_path`: The path to the checkpoint file to load from. Defaults to the `CHECKPOINT_PATH` constant,
  which is expected to be `"./input/model/v1-5-pruned-emaonly.safetensors"`.
- `-F, --fully_initialize`: Whether to fully initialize or not. Defaults to `False`.
- `-d, --distribution_index`: The distribution index to use. Defaults to 4. Options: 0: "Normal", 1: "Cauchy", 2: "
  Gumbel", 3: "Laplace", 4: "Logistic".
- `-s, --seed`: The seed value. Defaults to random int `0 to 2^24`.
- `-bs, --batch_size`: The batch size to use. Defaults to `1`.
- `--params_steps`: The number of steps for the parameters. Defaults to `3`.
- `--params_range`: The range of parameters. Defaults to `[0.49, 0.54]`.
- `--temperature_steps`: The number of steps for the temperature. Defaults to `3`.
- `--temperature_range`: The range of temperature. Defaults to `[1.0, 4.0]`.
- `--ddim_eta`: The value of ddim_eta. Defaults to `0.1`.
- `--clear_output_dir`: Whether to clear the output directory or not. Defaults to `False`.
- `--cuda_device`: The CUDA device to use. Defaults to `"get_device()"`.

Images and encodings

**Command line arguments**

- `--batch_size`: How many images to generate at once. Defaults to `1`.
- `--num_iterations`: How many times to iterate the generation of a batch of images. Defaults to `10`.
- `--prompt`: The prompt to render. It is an optional argument. Defaults
  to `"a painting of a cute monkey playing guitar"`.
- `--negative-prompt`: The negative prompt. For things we dont want to see in generated image. Defaults to `''`.
- `--cuda_device`: CUDA device to use for generation. Defaults to `"get_device()"`.
- `--low_vram`: Flag for low vram gpus. Defaults to `False`.
- `--sampler`: Sampler name. Defaults to `ddim`.
- `--cfg_scale`: Unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty)) . Defaults
  to `7`.
- `--seed`: Array of seed values, one for each generated image. Defaults to ``.

Try running:

python3 ./scripts/generate_images_and_encodings.py --prompt "An oil painting of a computer generated image of a geometric pattern" --num_iterations 10

Perturbations on prompts embeddings

Try running:

python3 ./scripts/embed_prompts_and_generate_images.py

Outputs in: ./output/disturbing_embeddings

• --prompt: The prompt to embed. Defaults to "A woman with flowers in her hair in a courtyard, in the style of Frank Frazetta".
• --num_iterations: The number of iterations to batch-generate images. Defaults to 8.
• --seed: The noise seed used to generate the images. Defaults to random int 0 to 2^24.
• --noise_multiplier: The multiplier for the amount of noise used to disturb the prompt embedding. Defaults to 0.01.
• --cuda_device: The CUDA device to use. Defaults to "get_device()".

Image Grid Generator

The Image Grid generator is a script that generates a grid of images from a directory or a zip file containing images.

Run the script with the following command:

python ./utility/scripts/grid_generator.py --input_path ./test/test_images/clip_segmentation --output_path ./tmp --rows 3 --columns 2  --img_size 256

Generate Images From Prompt Generator

To generate images from random prompt, these are the available CLI arguments:

options:
  -h, --help            show this help message and exit
  --batch_size BATCH_SIZE
                        How many images to generate at once (default: 1)
  --output OUTPUT       Path to the output directory (default: /output)
  --sampler SAMPLER     Name of the sampler to use (default: ddim)
  --checkpoint_path CHECKPOINT_PATH
                        Path to the checkpoint file (default: './input/model/v1-5-pruned-emaonly.safetensors')
  --flash               whether to use flash attention
  --steps STEPS         Number of steps to use (default: 20)
  --cfg_scale CFG_SCALE
                        unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty))
  --low_vram            limit VRAM usage
  --force_cpu           force CPU usage
  --cuda_device CUDA_DEVICE
                        cuda device to use for generation
  --num_images NUM_IMAGES
                        How many images to generate (default: 1)
  --seed SEED           Seed for the image generation (default: )
  --output_metadata     outputs the metadata
  --image_width IMAGE_WIDTH
                        Generate image width (default: 512)
  --image_height IMAGE_HEIGHT
                        Generate image height (default: 512)
  --num_datasets NUM_DATASETS
                        Number of datasets to generate (default: 1)
  --num-phrases [NUM_PHRASES]
                        The number of phrases for the prompt to generate

python3 ./scripts/generate_images_from_prompt_generator.py --checkpoint_path "./input/model/sd/v1-5-pruned-emaonly/v1-5-pruned-emaonly.safetensors" --cfg_scale 7 --num_images 10 --num_phrases 12 --output "./output/"

Generate Images From Prompt List Dataset

To generate images from prompt list dataset, these are the available CLI arguments:

options:
  -h, --help            show this help message and exit
  --batch_size BATCH_SIZE
                        How many images to generate at once (default: 1)
  --output OUTPUT       Path to the output directory (default: ./output)
  --sampler SAMPLER     Name of the sampler to use (default: ddim)
  --checkpoint_path CHECKPOINT_PATH
                        Path to the checkpoint file (default: './input/model/v1-5-pruned-emaonly.safetensors')
  --flash               whether to use flash attention
  --steps STEPS         Number of steps to use (default: 20)
  --cfg_scale CFG_SCALE
                        unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty))
  --force_cpu           force CPU usage
  --cuda_device CUDA_DEVICE
                        cuda device to use for generation
  --num_images NUM_IMAGES
                        How many images to generate (default: 1)
  --seed SEED           Seed for the image generation (default: )
  --image_width IMAGE_WIDTH
                        Generate image width (default: 512)
  --image_height IMAGE_HEIGHT
                        Generate image height (default: 512)
  --num_datasets NUM_DATASETS
                        Number of datasets to generate (default: 1)
  --image_batch_size IMAGE_BATCH_SIZE
                        Number of batches (default: 1)
  --prompt_list_dataset_path PROMPT_LIST_DATASET_PATH
                        The path to prompt list dataset zip
  --include_negative_prompt INCLUDE_NEGATIVE_PROMPT
                        True if negative prompt embedding will be used for generating images

python3 ./scripts/generate_images_from_prompt_list.py --checkpoint_path "./input/model/sd/v1-5-pruned-emaonly/v1-5-pruned-emaonly.safetensors" --cfg_scale 7 --num_images 2 --output ./output/generated-dataset-from-prompt-list --prompt_list_dataset_path ./test/test_zip_files/prompt_list_civitai_2_test.zip --include_negative_prompt True

Chad Score

To get the chad score of an image, these are the available CLI arguments:

options:
  --image-path IMAGE_PATH
                        Path to the image to be scored
  --model-path MODEL_PATH
                        Path to the model used for scoring the image

Example Usage:

python3 ./scripts/chad_score.py --model-path="input/model/chad_score/chad-score-v1.pth" --image-path="test/test_images/test_img.jpg"

Chad Sort

Takes in an image dataset Sorts the images in the dataset by chad score into multiple folders

These are the available CLI arguments:

options:
  --dataset-path DATASET_PATH
                        Path to the dataset to be sorted
  --output-path OUTPUT_PATH
                        Path to the output folder
  --num-classes NUM_CLASSES
                        Number of folders to output

Example Usage:

python3 ./scripts/chad_sort.py --dataset-path "test/test_zip_files/test-dataset-correct-format.zip" --output-path "/output/chad_sort/"

Running GenerationTask

Runs a generation task from .json file

These are the available CLI arguments:

options:
  --task_path TASK_PATH
                        Path to the task .json file

Example Usage:

python3 ./scripts/run_generation_task.py --task_path './test/test_generation_task/text_to_image_v1.json'

python3 ./scripts/run_generation_task.py --task_path './test/test_generation_task/generate_images_from_random_prompt_v1.json'

Prompt Score

Creates a linear regression model that uses prompt embedding tensors as input, and outputs a chad score.

options:
  --input_path INPUT_PATH
                        Path to input zip
  --output_path OUTPUT_PATH
                        Path output folder
  --model_output_name MODEL_OUTPUT_NAME
                        Filename of the trained model
  --num_epochs NUM_EPOCHS
                        Number of epochs (default: 1000)
  --epsilon_raw EPSILON_RAW
                        Epsilon for raw data (default: 10.0)
  --epsilon_scaled EPSILON_SCALED
                        Epsilon for scaled data (default: 0.2)
  --use_76th_embedding  If this option is set, only use the last entry in the embeddings tensor
  --show_validation_loss
                        whether to show validation loss

Example Usage:

python scripts/prompt_score.py --input_path input/set_0000_v2.zip --use_76th_embedding --num_epochs 200 --epsilon_raw 10 --epsilon_scaled 0.2 --model_output_name prompt_score.pth

Prompt Embeddings Gradient Optimization

Optimizes an embedding vector using gradients.

options:
  --input_path INPUT_PATH
                        Path to input zip
  --model_path MODEL_PATH
                        Path to the model
  --iterations ITERATIONS
                        How many iterations to perform
  --learning_rate LEARNING_RATE
                        Learning rate to use when optimizing

Example Usage:

python scripts/prompt_gradient.py --input_path input/set_0000_v2.zip --model_path output/models/prompt_score.pth --iterations 10 --learning_rate 0.01

Prompt Generator

Generates prompts and saves to a json file

usage: prompt_generator.py [-h] [--positive-prefix POSITIVE_PREFIX] [--num-prompts NUM_PROMPTS] [--csv-phrase-limit CSV_PHRASE_LIMIT] [--csv-path CSV_PATH] [--output OUTPUT] [--positive-ratio-threshold POSITIVE_RATIO_THRESHOLD] [--negative-ratio-threshold NEGATIVE_RATIO_THRESHOLD]
                           [--use-threshold USE_THRESHOLD] [--proportional-selection PROPORTIONAL_SELECTION]

Prompt Generator CLI tool generates prompts from phrases inside a csv

options:
  -h, --help            show this help message and exit
  --positive-prefix POSITIVE_PREFIX
                        Prefix phrase to add to positive prompts
  --num-prompts NUM_PROMPTS
                        Number of prompts to generate
  --csv-phrase-limit CSV_PHRASE_LIMIT
                        Number of phrases to use from the csv data
  --csv-path CSV_PATH   Full path to the csv path
  --output OUTPUT       Output path for dataset zip containing prompt list npz
  --positive-ratio-threshold POSITIVE_RATIO_THRESHOLD
                        Threshold ratio of positive/negative to use a phrase for positive prompt
  --negative-ratio-threshold NEGATIVE_RATIO_THRESHOLD
                        Threshold ratio of negative/positive to use a phrase for negative prompt
  --use-threshold USE_THRESHOLD
                        True if positive and negative ratio will be used
  --proportional-selection PROPORTIONAL_SELECTION
                        True if proportional selection will be used to get the phrases

To get civitai csv, download via:
https://mega.nz/file/wNIXHKDA#CIB92fLkCquatWvzUAjjcKbrKCUmO2ffRVrzE3XYQVM

Example Usage:

python ./scripts/prompt_generator.py --num-prompts 50 --positive-prefix "environmental, concept art, side scrolling, video game" --csv-phrase-limit 512 --csv-path ./input/civit_ai_data_phrase_count_v6.csv --output ./output/prompt_list_civitai_50_test --positive-ratio-threshold 3 --negative-ratio-threshold 3

python ./scripts/prompt_generator.py --num-prompts 50 --positive-prefix "environmental, concept art, side scrolling, video game" --csv-phrase-limit 512 --csv-path ./input/civit_ai_data_phrase_count_v6.csv --output ./output/prompt_list_civitai_50_test --proportional-selection True      

Image Ranker by Fitness Score

script will rank images according spedicif fitness functoin

Example Usage:

python3 scripts/fitness_score_ranker.py --fitness_function ga/fitness_bounding_box_centered.py --zip_path /path/to/the/zip/file --output_path /path/to/the/output

Auto-ml

Note: Current only support dataset generated from Generate Images Random Prompt

usage: auto_ml.py [-h] [--x-input X_INPUT] [--total-time TOTAL_TIME] [--per-run-time PER_RUN_TIME] [--dataset-zip-path DATASET_ZIP_PATH] [--output OUTPUT]

Run automl on a dataset with embeddings or clip feature and chad score

options:
  -h, --help            show this help message and exit
  --x-input X_INPUT     X-input will be either embedding or clip
  --total-time TOTAL_TIME
                        Time limit in seconds for the search of appropriate models
  --per-run-time PER_RUN_TIME
                        Time limit for a single call to the machine learning model
  --dataset-zip-path DATASET_ZIP_PATH
                        Path to the dataset to be used
  --output OUTPUT       Output path where the plot image will be saved

Example Usage:

python scripts/auto_ml.py --dataset-zip-path ./input/set_0002.zip --x-input "clip" --output "./output"

Generate Images From Model Predictions

Note: Currently only supports prediction output from kcg-ml/elm-regression

usage: generate_images_from_predictions.py [-h] [--dataset-path DATASET_PATH] [--output-path OUTPUT_PATH] [--num-class NUM_CLASS] [--limit-per-class LIMIT_PER_CLASS]

Generate Image For Each Range Of Predicted Chad Score

options:
  -h, --help            show this help message and exit
  --dataset-path DATASET_PATH
                        Predictions json path
  --output-path OUTPUT_PATH
                        Output path for the generated images
  --num-class NUM_CLASS
                        Number of classes to sort the images
  --limit-per-class LIMIT_PER_CLASS
                        Number of images to generate per class

Example Usage:

python ./scripts/generate_images_from_predictions.py --dataset-path "./output/chad-score-prediction.json" --output-path "./output/sort-generate-test" --num-class 10 --limit-per-class 1 

Split images

Split a bigger image example 512x512 into smaller pieces, example 64x64

options:
  -h, --help            show this help message and exit
  --image_path IMAGE_PATH
                        image input path
  --output   OUTPUT
                        Output path for the generated images
  --output_image_width OUTPUT_IMAGE_WIDTH
                        output image width
  --output_image_height OUTPUT_IMAGE_HEIGHT
                        output image height

Example Usage:

python scripts/split_image.py --image_path './test/test_images/512x512_test.png' --output './output/sub_images'

Affine Combination Of Embeddings GA

Combines N Number of prompts, the prompts are loaded from zip file Each prompt has a weight We search the weight vector space and try to maximize the fitness function The fitness function for now is chad score

options:
  -h, --help            show this help message and exit
  --generations GENERATIONS
                        The number of generations to execute
  --mutation_probability MUTATION_PROBABILITY
                        The probability of the occurence of a mutation
  --keep_elitism KEEP_ELITISM
                        Whether to keep the top fitness individual or not
  --crossover_type CROSSOVER_TYPE
                        Type of the crossover
  --mutation_type MUTATION_TYPE
                        Type of the mutation (swapn, random ...)
  --mutation_percent_genes MUTATION_PERCENT_GENES    
                        Percentage of the gene that will be mutated
  --population POPULATION
                        Number of starting population
  --steps STEPS
                        Number of steps for the sampler (iterations)
                       
  --device DEVICE
                        Defaults to cuda:0
  --num_phrases NUM_PHRASES
                        Number of phrases to use         
  --cfg_strength CFG_STRENGTH
                        Defaults to 12                 
  --sampler SAMPLER
                        Sampler name (ddim, ddpm)                  
                        
  --checkpoint_path CHECKPOINT_PATH
                        Path to the model checkpoint                  
                        
  --image_width IMAGE_WIDTH
                        Generated image width                
  --image_height IMAGE_HEIGHT
                        Generated image height                 
                        
  --output OUTPUT
                        Path to output folder                  
  --num_prompts NUM_PROMPTS
                        Number of prompts that will be used in the affine combination of embeddings                          
 --prompts_path PROMPTs_PATH
                        Path to the prompts zip file                        

Example Usage:

 python ./scripts/ga_affine_combination_embeddings.py --population 6 --num_prompts 6

Inpainting

Inpainting an image.

usage: inpaint_A1111.py [-h] --prompt PROMPT --init_img INIT_IMG --init_mask INIT_MASK [--negative_prompt NEGATIVE_PROMPT] [--sampler_name SAMPLER_NAME] [--batch_size BATCH_SIZE] [--n_iter N_ITER] [--steps STEPS] [--cfg_scale CFG_SCALE] [--width WIDTH] [--height HEIGHT]
                        [--mask_blur MASK_BLUR] [--inpainting_fill INPAINTING_FILL] [--outpath_samples OUTPATH_SAMPLES] [--styles STYLES [STYLES ...]] [--resize_mode RESIZE_MODE] [--denoising_strength DENOISING_STRENGTH]
                        [--image_cfg_scale IMAGE_CFG_SCALE] [--inpaint_full_res_padding INPAINT_FULL_RES_PADDING] [--inpainting_mask_invert INPAINTING_MASK_INVERT]

required arguments:
  --prompt PROMPT       Input prompt.
  --init_img INIT_IMG   Path to the initial image.
  --init_mask INIT_MASK Path to the initial mask.

options:
  -h, --help            show this help message and exit
  --negative_prompt NEGATIVE_PROMPT
                        Negative prompt.
  --sampler_name SAMPLER_NAME
                        Sampler name. Default: ddim.
  --batch_size BATCH_SIZE
                        Batch size. Default: 1.
  --n_iter N_ITER
                        Number of iterations. Default: 1.
  --steps STEPS
                        Steps. Default: 20.
  --cfg_scale CFG_SCALE
                        Config scale. Default: 7.0.
  --width WIDTH
                        Image width. Default: 512.
  --height HEIGHT
                        Image height. Default: 512.
  --mask_blur MASK_BLUR
                        Mask blur value. Default: 4.
  --inpainting_fill INPAINTING_FILL
                        Inpainting fill value. Default: 1.
  --outpath_samples OUTPATH_SAMPLES
                        Output path for samples. Default: output_samples_path.
  --styles STYLES [STYLES ...]
                        Styles list. Default: [].
  --resize_mode RESIZE_MODE
                        Resize mode. Default: 0.
  --denoising_strength DENOISING_STRENGTH
                        Denoising strength. Default: 0.75.
  --image_cfg_scale IMAGE_CFG_SCALE
                        Image config scale. Default: 1.5.
  --inpaint_full_res_padding INPAINT_FULL_RES_PADDING
                        Inpaint full resolution padding. Default: 32.
  --inpainting_mask_invert INPAINTING_MASK_INVERT
                        Inpainting mask invert value. Default: 0.

Exemple usage:

python ./scripts/inpaint_A1111.py --prompt "A cat paint by Vincent Van Gogh" --init_img "./test/test_inpainting/background.png" --init_mask "./test/test_inpainting/mask.png"

python ./scripts/inpaint_A1111.py --prompt "flat, vector art, game art, vector art, stylized, cell shaded, 8 bit, 16 bit, retro, russian futurism" --init_img "./test/test_inpainting/icon/background.png" --init_mask "./test/test_inpainting/icon/mask.png" --mask_blur 0 --n_iter 5 --metadata_output

Generate Images Using Inpainting from Prompt List Dataset

Generate image dataset using inpainting and prompts from prompt list dataset.

usage: generate_images_with_inpainting_from_prompt_list.py [-h] [--prompt_list_dataset_path PROMPT_LIST_DATASET_PATH] [--num_images NUM_IMAGES] [--init_img INIT_IMG] [--init_mask INIT_MASK] [--sampler_name SAMPLER_NAME] [--batch_size BATCH_SIZE] [--n_iter N_ITER] [--steps STEPS] [--cfg_scale CFG_SCALE]
                                                           [--width WIDTH] [--height HEIGHT] [--mask_blur MASK_BLUR] [--inpainting_fill INPAINTING_FILL] [--outpath OUTPATH] [--styles [STYLES ...]] [--resize_mode RESIZE_MODE] [--denoising_strength DENOISING_STRENGTH] [--image_cfg_scale IMAGE_CFG_SCALE]
                                                           [--inpaint_full_res_padding INPAINT_FULL_RES_PADDING] [--inpainting_mask_invert INPAINTING_MASK_INVERT]

required arguments:
  --prompt_list_dataset_path PROMPT_LIST_DATASET_PATH
                        The path of the prompt list dataset
  --num_images NUM_IMAGES
                        The number of images to generate
  --init_img INIT_IMG   Path to the initial image.
  --init_mask INIT_MASK Path to the initial mask.

options:
  -h, --help            show this help message and exit
  --sampler_name SAMPLER_NAME
                        Sampler name. Default: ddim.
  --batch_size BATCH_SIZE
                        Batch size. Default: 1.
  --n_iter N_ITER
                        Number of iterations. Default: 1.
  --steps STEPS
                        Steps. Default: 20.
  --cfg_scale CFG_SCALE
                        Config scale. Default: 7.0.
  --width WIDTH
                        Image width. Default: 512.
  --height HEIGHT
                        Image height. Default: 512.
  --mask_blur MASK_BLUR
                        Mask blur value. Default: 4.
  --inpainting_fill INPAINTING_FILL
                        Inpainting fill value. Default: 1.
  --outpath_samples OUTPATH_SAMPLES
                        Output path for samples. Default: output_samples_path.
  --styles STYLES [STYLES ...]
                        Styles list. Default: [].
  --resize_mode RESIZE_MODE
                        Resize mode. Default: 0.
  --denoising_strength DENOISING_STRENGTH
                        Denoising strength. Default: 0.75.
  --image_cfg_scale IMAGE_CFG_SCALE
                        Image config scale. Default: 1.5.
  --inpaint_full_res_padding INPAINT_FULL_RES_PADDING
                        Inpaint full resolution padding. Default: 32.
  --inpainting_mask_invert INPAINTING_MASK_INVERT
                        Inpainting mask invert value. Default: 0.

Exemple usage:

python ./scripts/generate_images_with_inpainting_from_prompt_list.py --init_img "./test/test_inpainting/white_512x512.jpg" --init_mask "./test/test_inpainting/character_mask.png" --prompt_list_dataset_path ./test/test_zip_files/prompt_list_civitai_2_test.zip --num_images 2

Tests

To run all scripts

1. Make an env

   python3 -m venv env

2. Activate env

   source env/bin/activate

3. Install requirements

   pip install -r requirements.txt

4. run pytest

   python -m unittest test_main_scripts.py

To run all kcg-ml-sd1p4 tests

1. Make an env

   python3 -m venv env

2. Activate env

   source env/bin/activate

3. Install requirements

   pip install -r requirements.txt

4. run pytest

   pytest ./test/test_scripts"