From: Christian Heller Date: Sun, 18 Aug 2024 23:56:23 +0000 (+0200) Subject: Reorganize code. X-Git-Url: https://plomlompom.com/repos/%7B%7Bprefix%7D%7D/static/pick_tasks?a=commitdiff_plain;h=0c5c4da0cb306ca35f49c55909c18268d67fc0ff;p=stable_plom Reorganize code. --- diff --git a/stable.py b/stable.py index 641ef48..ea0f598 100755 --- a/stable.py +++ b/stable.py @@ -2,9 +2,7 @@ from os.path import dirname, basename, splitext, join as path_join, exists from argparse import ArgumentParser from random import randint -from exiftool import ExifToolHelper # type: ignore -from torch import Generator -from stable.core import init_pipeline, make_metadata +from stable.core import ImageMaker def save_path(count: int) -> str: @@ -43,26 +41,15 @@ for n in range(args.number): if exists(path): raise Exception(f'Would overwrite file: {path}') -pipe = init_pipeline(args.model) -generator = Generator() +maker = ImageMaker(args.model) start_seed = args.randomness_seed start_seed = start_seed if start_seed != 0 else randint(-(2**31-1), 2**31) for n in range(args.number): nth_seed = start_seed + n path = save_path(n) - metadata = make_metadata(nth_seed, args.guidance, args.height, args.width, - args.model, args.prompt) - print(f'GENERATING: {path}; {metadata}') - generator.manual_seed(nth_seed) - images = pipe(args.prompt, - generator=generator, - guidance_scale=args.guidance, - num_inference_steps=args.steps, - height=args.height, - width=args.width - ).images - images[0].save(path) - with ExifToolHelper() as et: - et.set_tags([path], - tags={'Comment': metadata}, - params=['-overwrite_original']) + maker.set_gen_params(args.prompt, nth_seed, args.guidance, + args.height, args.width, args.steps) + print(f'GENERATING: {path}; {maker.gen_params_to_exif_comment}') + maker.set_gen_params(args.prompt, nth_seed, args.guidance, args.height, + args.width, args.steps) + maker.gen_image_to(path) diff --git a/stable/core.py b/stable/core.py index d7f9d23..5612ab2 100644 --- a/stable/core.py +++ b/stable/core.py @@ -1,36 +1,85 @@ -from logging import (Formatter as LoggingFormatter, captureWarnings, - Filter as LoggingFilter) +from logging import (Formatter as LogFormatter, captureWarnings, + Filter as LogFilter) +from os.path import basename from diffusers import StableDiffusionPipeline from diffusers.utils import logging +from torch import Generator +from exiftool import ExifToolHelper # type: ignore -LOG_FMT = 'DIFFUSERS WARNING: %(message)s\n' -SAFETY_MSG_PATTERN = 'You have disabled the safety checker' +SAFETY_CHECKER_WARNING_PATTERN = 'You have disabled the safety checker' -class _FilterOutString(LoggingFilter): +class ImageMaker: - def __init__(self, target): - super().__init__() - self.target = target + def __init__(self, model_path): - def filter(self, record): - return self.target not in record.getMessage() + class FilterOut(LogFilter): + def __init__(self, target): + super().__init__() + self.target = target -def init_pipeline(model): - print(f'SETTING UP STABLE DIFFUSION PIPELINE FROM MODEL: {model}\n') - diffusers_logging_handler = logging.get_logger('diffusers').handlers[0] - diffusers_logging_handler.setFormatter(LoggingFormatter(fmt=LOG_FMT)) - diffusers_logging_handler.addFilter(_FilterOutString(SAFETY_MSG_PATTERN)) - captureWarnings(True) - logging.disable_progress_bar() - pipe = StableDiffusionPipeline.from_single_file(model, - local_files_only=True) - pipe.to('cuda') - print('PIPELINE READY\n') - return pipe + def filter(self, record): + return self.target not in record.getMessage() + self.model_filename = basename(model_path) + self.seed = None + self.prompt = None + self.guidance = None + self.height = None + self.width = None + self.n_steps = None + prefix = 'SETTING UP STABLE DIFFUSION PIPELINE FROM MODEL' + print(f'{prefix}: {model_path}\n') + diffusers_logging_handler = logging.get_logger('diffusers').handlers[0] + diffusers_logging_handler.setFormatter( + LogFormatter(fmt='DIFFUSERS WARNING: %(message)s\n')) + diffusers_logging_handler.addFilter( + FilterOut(SAFETY_CHECKER_WARNING_PATTERN)) + captureWarnings(True) + logging.disable_progress_bar() + self.pipe = StableDiffusionPipeline.from_single_file( + model_path, local_files_only=True) + self.pipe.to('cuda') + self.generator = Generator() + print('PIPELINE READY\n') -def make_metadata(seed, guidance, height, width, model, prompt): - return f'SEED: {seed}; GUIDANCE: {guidance}; HEIGHT: {height}; ' +\ - f'WIDTH: {width}; MODEL: {model}; PROMPT: {prompt}' + def set_seed(self, seed): + self.seed = seed + self.generator.manual_seed(seed) + + def set_gen_params(self, prompt, seed, guidance, height, width, n_steps): + self.seed = seed + self.prompt = prompt + self.guidance = guidance + self.height = height + self.width = width + self.n_steps = n_steps + + @property + def gen_params_to_exif_comment(self): + return f'SEED: {self.seed}; ' +\ + f'GUIDANCE: {self.guidance}; ' +\ + f'NUMBER OF STEPS: {self.n_steps}; ' +\ + f'HEIGHT: {self.height}; ' +\ + f'WIDTH: {self.width}; ' +\ + f'MODEL_FILE: {self.model_filename}; ' +\ + f'PROMPT: {self.prompt}' + + def gen_image_to(self, path): + if None in {self.seed, self.prompt, self.guidance, self.height, + self.width, self.n_steps}: + raise Exception('Generation parameters not initialized.') + self.generator.manual_seed(self.seed) + image = self.pipe(generator=self.generator, + prompt=self.prompt, + guidance=self.guidance, + height=self.height, + width=self.width, + num_inference_steps=self.n_steps, + ).images[0] + image.save(path) + with ExifToolHelper() as et: + et.set_tags([path], + tags={'Comment': self.gen_params_to_exif_comment}, + params=['-overwrite_original'])