Foreign-Mirrors/Fooocus
1
0
Fork 0
mirror of https://github.com/lllyasviel/Fooocus.git synced 2026-08-16 13:13:16 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'hotfix/prevent-skipping-and-stopping-by-other-users'

Browse Source 
# Conflicts:
#	modules/advanced_parameters.py
#	modules/async_worker.py
#	webui.py
This commit is contained in:
Manuel Schmid
2024-01-22 23:15:18 +01:00
parent 25c157b177 148eddf48d
commit d7c1f4a6aa
9 changed files with 222 additions and 208 deletions
Download Patch File Download Diff File Expand all files Collapse all files
extras/preprocessors.py
+8 -9
View File
@@ -1,27 +1,26 @@
import cv2 import cv2
import numpy as np import numpy as np
import modules.advanced_parameters as advanced_parameters
def centered_canny(x: np.ndarray): def centered_canny(x: np.ndarray, canny_low_threshold, canny_high_threshold):
assert isinstance(x, np.ndarray) assert isinstance(x, np.ndarray)
assert x.ndim == 2 and x.dtype == np.uint8 assert x.ndim == 2 and x.dtype == np.uint8
y = cv2.Canny(x, int(advanced_parameters.canny_low_threshold), int(advanced_parameters.canny_high_threshold)) y = cv2.Canny(x, int(canny_low_threshold), int(canny_high_threshold))
y = y.astype(np.float32) / 255.0 y = y.astype(np.float32) / 255.0
return y return y
def centered_canny_color(x: np.ndarray): def centered_canny_color(x: np.ndarray, canny_low_threshold, canny_high_threshold):
assert isinstance(x, np.ndarray) assert isinstance(x, np.ndarray)
assert x.ndim == 3 and x.shape[2] == 3 assert x.ndim == 3 and x.shape[2] == 3
result = [centered_canny(x[..., i]) for i in range(3)] result = [centered_canny(x[..., i], canny_low_threshold, canny_high_threshold) for i in range(3)]
result = np.stack(result, axis=2) result = np.stack(result, axis=2)
return result return result
def pyramid_canny_color(x: np.ndarray): def pyramid_canny_color(x: np.ndarray, canny_low_threshold, canny_high_threshold):
assert isinstance(x, np.ndarray) assert isinstance(x, np.ndarray)
assert x.ndim == 3 and x.shape[2] == 3 assert x.ndim == 3 and x.shape[2] == 3
@@ -31,7 +30,7 @@ def pyramid_canny_color(x: np.ndarray):
for k in [0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]: for k in [0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]:
Hs, Ws = int(H * k), int(W * k) Hs, Ws = int(H * k), int(W * k)
small = cv2.resize(x, (Ws, Hs), interpolation=cv2.INTER_AREA) small = cv2.resize(x, (Ws, Hs), interpolation=cv2.INTER_AREA)
edge = centered_canny_color(small) edge = centered_canny_color(small, canny_low_threshold, canny_high_threshold)
if acc_edge is None: if acc_edge is None:
acc_edge = edge acc_edge = edge
else: else:
@@ -54,11 +53,11 @@ def norm255(x, low=4, high=96):
return x * 255.0 return x * 255.0
def canny_pyramid(x): def canny_pyramid(x, canny_low_threshold, canny_high_threshold):
# For some reasons, SAI's Control-lora Canny seems to be trained on canny maps with non-standard resolutions. # For some reasons, SAI's Control-lora Canny seems to be trained on canny maps with non-standard resolutions.
# Then we use pyramid to use all resolutions to avoid missing any structure in specific resolutions. # Then we use pyramid to use all resolutions to avoid missing any structure in specific resolutions.
color_canny = pyramid_canny_color(x) color_canny = pyramid_canny_color(x, canny_low_threshold, canny_high_threshold)
result = np.sum(color_canny, axis=2) result = np.sum(color_canny, axis=2)
return norm255(result, low=1, high=99).clip(0, 255).astype(np.uint8) return norm255(result, low=1, high=99).clip(0, 255).astype(np.uint8)
modules/advanced_parameters.py
-32
View File
@@ -1,32 +0,0 @@
disable_preview, disable_intermediate_results, black_out_nsfw, adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg, sampler_name, \
scheduler_name, generate_image_grid, overwrite_step, overwrite_switch, overwrite_width, overwrite_height, \
overwrite_vary_strength, overwrite_upscale_strength, \
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint, \
debugging_cn_preprocessor, skipping_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold, \
refiner_swap_method, \
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2, \
debugging_inpaint_preprocessor, inpaint_disable_initial_latent, inpaint_engine, inpaint_strength, inpaint_respective_field, \
inpaint_mask_upload_checkbox, invert_mask_checkbox, inpaint_erode_or_dilate = [None] * 37
def set_all_advanced_parameters(*args):
global disable_preview, disable_intermediate_results, black_out_nsfw, adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg, sampler_name, \
scheduler_name, generate_image_grid, overwrite_step, overwrite_switch, overwrite_width, overwrite_height, \
overwrite_vary_strength, overwrite_upscale_strength, \
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint, \
debugging_cn_preprocessor, skipping_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold, \
refiner_swap_method, \
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2, \
debugging_inpaint_preprocessor, inpaint_disable_initial_latent, inpaint_engine, inpaint_strength, inpaint_respective_field, \
inpaint_mask_upload_checkbox, invert_mask_checkbox, inpaint_erode_or_dilate
disable_preview, disable_intermediate_results, black_out_nsfw, adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg, sampler_name, \
scheduler_name, generate_image_grid, overwrite_step, overwrite_switch, overwrite_width, overwrite_height, \
overwrite_vary_strength, overwrite_upscale_strength, \
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint, \
debugging_cn_preprocessor, skipping_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold, \
refiner_swap_method, \
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2, \
debugging_inpaint_preprocessor, inpaint_disable_initial_latent, inpaint_engine, inpaint_strength, inpaint_respective_field, \
inpaint_mask_upload_checkbox, invert_mask_checkbox, inpaint_erode_or_dilate = args
return
modules/async_worker.py
+144 -96
View File
@@ -1,4 +1,8 @@
import threading import threading
import os
from modules.patch import PatchSettings, patch_settings, patch_all
patch_all()
class AsyncTask: class AsyncTask:
@@ -35,7 +39,6 @@ def worker():
import extras.preprocessors as preprocessors import extras.preprocessors as preprocessors
import modules.inpaint_worker as inpaint_worker import modules.inpaint_worker as inpaint_worker
import modules.constants as constants import modules.constants as constants
import modules.advanced_parameters as advanced_parameters
import extras.ip_adapter as ip_adapter import extras.ip_adapter as ip_adapter
import extras.face_crop import extras.face_crop
import fooocus_version import fooocus_version
@@ -50,6 +53,9 @@ def worker():
get_image_shape_ceil, set_image_shape_ceil, get_shape_ceil, resample_image, erode_or_dilate, calculate_sha256, quote get_image_shape_ceil, set_image_shape_ceil, get_shape_ceil, resample_image, erode_or_dilate, calculate_sha256, quote
from modules.upscaler import perform_upscale from modules.upscaler import perform_upscale
pid = os.getpid()
print(f'Started worker with PID {pid}')
try: try:
async_gradio_app = shared.gradio_root async_gradio_app = shared.gradio_root
flag = f'''App started successful. Use the app with {str(async_gradio_app.local_url)} or {str(async_gradio_app.server_name)}:{str(async_gradio_app.server_port)}''' flag = f'''App started successful. Use the app with {str(async_gradio_app.local_url)} or {str(async_gradio_app.server_name)}:{str(async_gradio_app.server_port)}'''
@@ -63,14 +69,10 @@ def worker():
print(f'[Fooocus] {text}') print(f'[Fooocus] {text}')
async_task.yields.append(['preview', (number, text, None)]) async_task.yields.append(['preview', (number, text, None)])
def yield_result(async_task, imgs, do_not_show_finished_images=False, progressbar_index=13): def yield_result(async_task, imgs, black_out_nsfw, do_not_show_finished_images=False):
if not isinstance(imgs, list): if not isinstance(imgs, list):
imgs = [imgs] imgs = [imgs]
if modules.config.default_black_out_nsfw or advanced_parameters.black_out_nsfw:
progressbar(async_task, progressbar_index, 'Checking for NSFW content ...')
imgs = censor_batch(imgs)
async_task.results = async_task.results + imgs async_task.results = async_task.results + imgs
if do_not_show_finished_images: if do_not_show_finished_images:
@@ -80,9 +82,6 @@ def worker():
return return
def build_image_wall(async_task): def build_image_wall(async_task):
if not advanced_parameters.generate_image_grid:
return
results = async_task.results results = async_task.results
if len(results) < 2: if len(results) < 2:
@@ -154,6 +153,44 @@ def worker():
inpaint_input_image = args.pop() inpaint_input_image = args.pop()
inpaint_additional_prompt = args.pop() inpaint_additional_prompt = args.pop()
inpaint_mask_image_upload = args.pop() inpaint_mask_image_upload = args.pop()
disable_preview = args.pop()
disable_intermediate_results = args.pop()
black_out_nsfw = args.pop()
adm_scaler_positive = args.pop()
adm_scaler_negative = args.pop()
adm_scaler_end = args.pop()
adaptive_cfg = args.pop()
sampler_name = args.pop()
scheduler_name = args.pop()
overwrite_step = args.pop()
overwrite_switch = args.pop()
overwrite_width = args.pop()
overwrite_height = args.pop()
overwrite_vary_strength = args.pop()
overwrite_upscale_strength = args.pop()
mixing_image_prompt_and_vary_upscale = args.pop()
mixing_image_prompt_and_inpaint = args.pop()
debugging_cn_preprocessor = args.pop()
skipping_cn_preprocessor = args.pop()
canny_low_threshold = args.pop()
canny_high_threshold = args.pop()
refiner_swap_method = args.pop()
controlnet_softness = args.pop()
freeu_enabled = args.pop()
freeu_b1 = args.pop()
freeu_b2 = args.pop()
freeu_s1 = args.pop()
freeu_s2 = args.pop()
debugging_inpaint_preprocessor = args.pop()
inpaint_disable_initial_latent = args.pop()
inpaint_engine = args.pop()
inpaint_strength = args.pop()
inpaint_respective_field = args.pop()
inpaint_mask_upload_checkbox = args.pop()
invert_mask_checkbox = args.pop()
inpaint_erode_or_dilate = args.pop()
save_metadata_to_images = args.pop() if not args_manager.args.disable_metadata else False save_metadata_to_images = args.pop() if not args_manager.args.disable_metadata else False
metadata_scheme = args.pop() if not args_manager.args.disable_metadata else 'fooocus' metadata_scheme = args.pop() if not args_manager.args.disable_metadata else 'fooocus'
@@ -202,15 +239,15 @@ def worker():
print(f'Refiner disabled in LCM mode.') print(f'Refiner disabled in LCM mode.')
refiner_model_name = 'None' refiner_model_name = 'None'
sampler_name = advanced_parameters.sampler_name = 'lcm' sampler_name = 'lcm'
scheduler_name = advanced_parameters.scheduler_name = 'lcm' scheduler_name = 'lcm'
modules.patch.sharpness = sharpness = 0.0 sharpness = 0.0
cfg_scale = guidance_scale = 1.0 guidance_scale = 1.0
modules.patch.adaptive_cfg = advanced_parameters.adaptive_cfg = 1.0 adaptive_cfg = 1.0
refiner_switch = 1.0 refiner_switch = 1.0
modules.patch.positive_adm_scale = advanced_parameters.adm_scaler_positive = 1.0 adm_scaler_positive = 1.0
modules.patch.negative_adm_scale = advanced_parameters.adm_scaler_negative = 1.0 adm_scaler_negative = 1.0
modules.patch.adm_scaler_end = advanced_parameters.adm_scaler_end = 0.0 adm_scaler_end = 0.0
steps = 8 steps = 8
if translate_prompts: if translate_prompts:
@@ -229,19 +266,22 @@ def worker():
lora_hashes.append(f'{n.split(".")[0]}: {calculate_sha256(lora_path)[0:10]}') lora_hashes.append(f'{n.split(".")[0]}: {calculate_sha256(lora_path)[0:10]}')
lora_hashes_string = ", ".join(lora_hashes) lora_hashes_string = ", ".join(lora_hashes)
modules.patch.adaptive_cfg = advanced_parameters.adaptive_cfg print(f'[Parameters] Adaptive CFG = {adaptive_cfg}')
print(f'[Parameters] Adaptive CFG = {modules.patch.adaptive_cfg}') print(f'[Parameters] Sharpness = {sharpness}')
print(f'[Parameters] ControlNet Softness = {controlnet_softness}')
modules.patch.sharpness = sharpness
print(f'[Parameters] Sharpness = {modules.patch.sharpness}')
modules.patch.positive_adm_scale = advanced_parameters.adm_scaler_positive
modules.patch.negative_adm_scale = advanced_parameters.adm_scaler_negative
modules.patch.adm_scaler_end = advanced_parameters.adm_scaler_end
print(f'[Parameters] ADM Scale = ' print(f'[Parameters] ADM Scale = '
f'{modules.patch.positive_adm_scale} : ' f'{adm_scaler_positive} : '
f'{modules.patch.negative_adm_scale} : ' f'{adm_scaler_negative} : '
f'{modules.patch.adm_scaler_end}') f'{adm_scaler_end}')
patch_settings[pid] = PatchSettings(
sharpness,
adm_scaler_end,
adm_scaler_positive,
adm_scaler_negative,
controlnet_softness,
adaptive_cfg
)
cfg_scale = float(guidance_scale) cfg_scale = float(guidance_scale)
print(f'[Parameters] CFG = {cfg_scale}') print(f'[Parameters] CFG = {cfg_scale}')
@@ -254,10 +294,9 @@ def worker():
width, height = int(width), int(height) width, height = int(width), int(height)
skip_prompt_processing = False skip_prompt_processing = False
refiner_swap_method = advanced_parameters.refiner_swap_method
inpaint_worker.current_task = None inpaint_worker.current_task = None
inpaint_parameterized = advanced_parameters.inpaint_engine != 'None' inpaint_parameterized = inpaint_engine != 'None'
inpaint_image = None inpaint_image = None
inpaint_mask = None inpaint_mask = None
inpaint_head_model_path = None inpaint_head_model_path = None
@@ -271,15 +310,12 @@ def worker():
seed = int(image_seed) seed = int(image_seed)
print(f'[Parameters] Seed = {seed}') print(f'[Parameters] Seed = {seed}')
sampler_name = advanced_parameters.sampler_name
scheduler_name = advanced_parameters.scheduler_name
goals = [] goals = []
tasks = [] tasks = []
if input_image_checkbox: if input_image_checkbox:
if (current_tab == 'uov' or ( if (current_tab == 'uov' or (
current_tab == 'ip' and advanced_parameters.mixing_image_prompt_and_vary_upscale)) \ current_tab == 'ip' and mixing_image_prompt_and_vary_upscale)) \
and uov_method != flags.disabled and uov_input_image is not None: and uov_method != flags.disabled and uov_input_image is not None:
uov_input_image = HWC3(uov_input_image) uov_input_image = HWC3(uov_input_image)
if 'vary' in uov_method: if 'vary' in uov_method:
@@ -303,12 +339,12 @@ def worker():
progressbar(async_task, 1, 'Downloading upscale models ...') progressbar(async_task, 1, 'Downloading upscale models ...')
modules.config.downloading_upscale_model() modules.config.downloading_upscale_model()
if (current_tab == 'inpaint' or ( if (current_tab == 'inpaint' or (
current_tab == 'ip' and advanced_parameters.mixing_image_prompt_and_inpaint)) \ current_tab == 'ip' and mixing_image_prompt_and_inpaint)) \
and isinstance(inpaint_input_image, dict): and isinstance(inpaint_input_image, dict):
inpaint_image = inpaint_input_image['image'] inpaint_image = inpaint_input_image['image']
inpaint_mask = inpaint_input_image['mask'][:, :, 0] inpaint_mask = inpaint_input_image['mask'][:, :, 0]
if advanced_parameters.inpaint_mask_upload_checkbox: if inpaint_mask_upload_checkbox:
if isinstance(inpaint_mask_image_upload, np.ndarray): if isinstance(inpaint_mask_image_upload, np.ndarray):
if inpaint_mask_image_upload.ndim == 3: if inpaint_mask_image_upload.ndim == 3:
H, W, C = inpaint_image.shape H, W, C = inpaint_image.shape
@@ -317,10 +353,10 @@ def worker():
inpaint_mask_image_upload = (inpaint_mask_image_upload > 127).astype(np.uint8) * 255 inpaint_mask_image_upload = (inpaint_mask_image_upload > 127).astype(np.uint8) * 255
inpaint_mask = np.maximum(inpaint_mask, inpaint_mask_image_upload) inpaint_mask = np.maximum(inpaint_mask, inpaint_mask_image_upload)
if int(advanced_parameters.inpaint_erode_or_dilate) != 0: if int(inpaint_erode_or_dilate) != 0:
inpaint_mask = erode_or_dilate(inpaint_mask, advanced_parameters.inpaint_erode_or_dilate) inpaint_mask = erode_or_dilate(inpaint_mask, inpaint_erode_or_dilate)
if advanced_parameters.invert_mask_checkbox: if invert_mask_checkbox:
inpaint_mask = 255 - inpaint_mask inpaint_mask = 255 - inpaint_mask
inpaint_image = HWC3(inpaint_image) inpaint_image = HWC3(inpaint_image)
@@ -331,7 +367,7 @@ def worker():
if inpaint_parameterized: if inpaint_parameterized:
progressbar(async_task, 1, 'Downloading inpainter ...') progressbar(async_task, 1, 'Downloading inpainter ...')
inpaint_head_model_path, inpaint_patch_model_path = modules.config.downloading_inpaint_models( inpaint_head_model_path, inpaint_patch_model_path = modules.config.downloading_inpaint_models(
advanced_parameters.inpaint_engine) inpaint_engine)
base_model_additional_loras += [(inpaint_patch_model_path, 1.0)] base_model_additional_loras += [(inpaint_patch_model_path, 1.0)]
print(f'[Inpaint] Current inpaint model is {inpaint_patch_model_path}') print(f'[Inpaint] Current inpaint model is {inpaint_patch_model_path}')
if refiner_model_name == 'None': if refiner_model_name == 'None':
@@ -347,8 +383,8 @@ def worker():
prompt = inpaint_additional_prompt + '\n' + prompt prompt = inpaint_additional_prompt + '\n' + prompt
goals.append('inpaint') goals.append('inpaint')
if current_tab == 'ip' or \ if current_tab == 'ip' or \
advanced_parameters.mixing_image_prompt_and_inpaint or \ mixing_image_prompt_and_vary_upscale or \
advanced_parameters.mixing_image_prompt_and_vary_upscale: mixing_image_prompt_and_inpaint:
goals.append('cn') goals.append('cn')
progressbar(async_task, 1, 'Downloading control models ...') progressbar(async_task, 1, 'Downloading control models ...')
if len(cn_tasks[flags.cn_canny]) > 0: if len(cn_tasks[flags.cn_canny]) > 0:
@@ -369,17 +405,17 @@ def worker():
switch = int(round(steps * refiner_switch)) switch = int(round(steps * refiner_switch))
if advanced_parameters.overwrite_step > 0: if overwrite_step > 0:
steps = advanced_parameters.overwrite_step steps = overwrite_step
if advanced_parameters.overwrite_switch > 0: if overwrite_switch > 0:
switch = advanced_parameters.overwrite_switch switch = overwrite_switch
if advanced_parameters.overwrite_width > 0: if overwrite_width > 0:
width = advanced_parameters.overwrite_width width = overwrite_width
if advanced_parameters.overwrite_height > 0: if overwrite_height > 0:
height = advanced_parameters.overwrite_height height = overwrite_height
print(f'[Parameters] Sampler = {sampler_name} - {scheduler_name}') print(f'[Parameters] Sampler = {sampler_name} - {scheduler_name}')
print(f'[Parameters] Steps = {steps} - {switch}') print(f'[Parameters] Steps = {steps} - {switch}')
@@ -482,8 +518,8 @@ def worker():
denoising_strength = 0.5 denoising_strength = 0.5
if 'strong' in uov_method: if 'strong' in uov_method:
denoising_strength = 0.85 denoising_strength = 0.85
if advanced_parameters.overwrite_vary_strength > 0: if overwrite_vary_strength > 0:
denoising_strength = advanced_parameters.overwrite_vary_strength denoising_strength = overwrite_vary_strength
shape_ceil = get_image_shape_ceil(uov_input_image) shape_ceil = get_image_shape_ceil(uov_input_image)
if shape_ceil < 1024: if shape_ceil < 1024:
@@ -548,14 +584,14 @@ def worker():
if direct_return: if direct_return:
d = [('Upscale (Fast)', '2x')] d = [('Upscale (Fast)', '2x')]
uov_input_image_path = log(uov_input_image, d, output_format=output_format) uov_input_image_path = log(uov_input_image, d, output_format=output_format)
yield_result(async_task, uov_input_image_path, do_not_show_finished_images=True) yield_result(async_task, uov_input_image_path, black_out_nsfw, do_not_show_finished_images=True)
return return
tiled = True tiled = True
denoising_strength = 0.382 denoising_strength = 0.382
if advanced_parameters.overwrite_upscale_strength > 0: if overwrite_upscale_strength > 0:
denoising_strength = advanced_parameters.overwrite_upscale_strength denoising_strength = overwrite_upscale_strength
initial_pixels = core.numpy_to_pytorch(uov_input_image) initial_pixels = core.numpy_to_pytorch(uov_input_image)
progressbar(async_task, 13, 'VAE encoding ...') progressbar(async_task, 13, 'VAE encoding ...')
@@ -599,20 +635,20 @@ def worker():
inpaint_image = np.ascontiguousarray(inpaint_image.copy()) inpaint_image = np.ascontiguousarray(inpaint_image.copy())
inpaint_mask = np.ascontiguousarray(inpaint_mask.copy()) inpaint_mask = np.ascontiguousarray(inpaint_mask.copy())
advanced_parameters.inpaint_strength = 1.0 inpaint_strength = 1.0
advanced_parameters.inpaint_respective_field = 1.0 inpaint_respective_field = 1.0
denoising_strength = advanced_parameters.inpaint_strength denoising_strength = inpaint_strength
inpaint_worker.current_task = inpaint_worker.InpaintWorker( inpaint_worker.current_task = inpaint_worker.InpaintWorker(
image=inpaint_image, image=inpaint_image,
mask=inpaint_mask, mask=inpaint_mask,
use_fill=denoising_strength > 0.99, use_fill=denoising_strength > 0.99,
k=advanced_parameters.inpaint_respective_field k=inpaint_respective_field
) )
if advanced_parameters.debugging_inpaint_preprocessor: if debugging_inpaint_preprocessor:
yield_result(async_task, inpaint_worker.current_task.visualize_mask_processing(), yield_result(async_task, inpaint_worker.current_task.visualize_mask_processing(), black_out_nsfw,
do_not_show_finished_images=True) do_not_show_finished_images=True)
return return
@@ -657,7 +693,7 @@ def worker():
model=pipeline.final_unet model=pipeline.final_unet
) )
if not advanced_parameters.inpaint_disable_initial_latent: if not inpaint_disable_initial_latent:
initial_latent = {'samples': latent_fill} initial_latent = {'samples': latent_fill}
B, C, H, W = latent_fill.shape B, C, H, W = latent_fill.shape
@@ -670,25 +706,25 @@ def worker():
cn_img, cn_stop, cn_weight = task cn_img, cn_stop, cn_weight = task
cn_img = resize_image(HWC3(cn_img), width=width, height=height) cn_img = resize_image(HWC3(cn_img), width=width, height=height)
if not advanced_parameters.skipping_cn_preprocessor: if not skipping_cn_preprocessor:
cn_img = preprocessors.canny_pyramid(cn_img) cn_img = preprocessors.canny_pyramid(cn_img, canny_low_threshold, canny_high_threshold)
cn_img = HWC3(cn_img) cn_img = HWC3(cn_img)
task[0] = core.numpy_to_pytorch(cn_img) task[0] = core.numpy_to_pytorch(cn_img)
if advanced_parameters.debugging_cn_preprocessor: if debugging_cn_preprocessor:
yield_result(async_task, cn_img, do_not_show_finished_images=True) yield_result(async_task, cn_img, black_out_nsfw, do_not_show_finished_images=True)
return return
for task in cn_tasks[flags.cn_cpds]: for task in cn_tasks[flags.cn_cpds]:
cn_img, cn_stop, cn_weight = task cn_img, cn_stop, cn_weight = task
cn_img = resize_image(HWC3(cn_img), width=width, height=height) cn_img = resize_image(HWC3(cn_img), width=width, height=height)
if not advanced_parameters.skipping_cn_preprocessor: if not skipping_cn_preprocessor:
cn_img = preprocessors.cpds(cn_img) cn_img = preprocessors.cpds(cn_img)
cn_img = HWC3(cn_img) cn_img = HWC3(cn_img)
task[0] = core.numpy_to_pytorch(cn_img) task[0] = core.numpy_to_pytorch(cn_img)
if advanced_parameters.debugging_cn_preprocessor: if debugging_cn_preprocessor:
yield_result(async_task, cn_img, do_not_show_finished_images=True) yield_result(async_task, cn_img, black_out_nsfw, do_not_show_finished_images=True)
return return
for task in cn_tasks[flags.cn_ip]: for task in cn_tasks[flags.cn_ip]:
cn_img, cn_stop, cn_weight = task cn_img, cn_stop, cn_weight = task
@@ -698,22 +734,22 @@ def worker():
cn_img = resize_image(cn_img, width=224, height=224, resize_mode=0) cn_img = resize_image(cn_img, width=224, height=224, resize_mode=0)
task[0] = ip_adapter.preprocess(cn_img, ip_adapter_path=ip_adapter_path) task[0] = ip_adapter.preprocess(cn_img, ip_adapter_path=ip_adapter_path)
if advanced_parameters.debugging_cn_preprocessor: if debugging_cn_preprocessor:
yield_result(async_task, cn_img, do_not_show_finished_images=True) yield_result(async_task, cn_img, black_out_nsfw, do_not_show_finished_images=True)
return return
for task in cn_tasks[flags.cn_ip_face]: for task in cn_tasks[flags.cn_ip_face]:
cn_img, cn_stop, cn_weight = task cn_img, cn_stop, cn_weight = task
cn_img = HWC3(cn_img) cn_img = HWC3(cn_img)
if not advanced_parameters.skipping_cn_preprocessor: if not skipping_cn_preprocessor:
cn_img = extras.face_crop.crop_image(cn_img) cn_img = extras.face_crop.crop_image(cn_img)
# https://github.com/tencent-ailab/IP-Adapter/blob/d580c50a291566bbf9fc7ac0f760506607297e6d/README.md?plain=1#L75 # https://github.com/tencent-ailab/IP-Adapter/blob/d580c50a291566bbf9fc7ac0f760506607297e6d/README.md?plain=1#L75
cn_img = resize_image(cn_img, width=224, height=224, resize_mode=0) cn_img = resize_image(cn_img, width=224, height=224, resize_mode=0)
task[0] = ip_adapter.preprocess(cn_img, ip_adapter_path=ip_adapter_face_path) task[0] = ip_adapter.preprocess(cn_img, ip_adapter_path=ip_adapter_face_path)
if advanced_parameters.debugging_cn_preprocessor: if debugging_cn_preprocessor:
yield_result(async_task, cn_img, do_not_show_finished_images=True) yield_result(async_task, cn_img, black_out_nsfw, do_not_show_finished_images=True)
return return
all_ip_tasks = cn_tasks[flags.cn_ip] + cn_tasks[flags.cn_ip_face] all_ip_tasks = cn_tasks[flags.cn_ip] + cn_tasks[flags.cn_ip_face]
@@ -721,14 +757,14 @@ def worker():
if len(all_ip_tasks) > 0: if len(all_ip_tasks) > 0:
pipeline.final_unet = ip_adapter.patch_model(pipeline.final_unet, all_ip_tasks) pipeline.final_unet = ip_adapter.patch_model(pipeline.final_unet, all_ip_tasks)
if advanced_parameters.freeu_enabled: if freeu_enabled:
print(f'FreeU is enabled!') print(f'FreeU is enabled!')
pipeline.final_unet = core.apply_freeu( pipeline.final_unet = core.apply_freeu(
pipeline.final_unet, pipeline.final_unet,
advanced_parameters.freeu_b1, freeu_b1,
advanced_parameters.freeu_b2, freeu_b2,
advanced_parameters.freeu_s1, freeu_s1,
advanced_parameters.freeu_s2 freeu_s2
) )
all_steps = steps * image_number all_steps = steps * image_number
@@ -804,7 +840,8 @@ def worker():
denoise=denoising_strength, denoise=denoising_strength,
tiled=tiled, tiled=tiled,
cfg_scale=cfg_scale, cfg_scale=cfg_scale,
refiner_swap_method=refiner_swap_method refiner_swap_method=refiner_swap_method,
disable_preview=disable_preview
) )
del task['c'], task['uc'], positive_cond, negative_cond # Save memory del task['c'], task['uc'], positive_cond, negative_cond # Save memory
@@ -823,15 +860,14 @@ def worker():
'steps': steps, 'refiner_switch': refiner_switch, 'sharpness': sharpness, 'cfg': cfg_scale, 'steps': steps, 'refiner_switch': refiner_switch, 'sharpness': sharpness, 'cfg': cfg_scale,
'base_model': base_model_name, 'refiner_model': refiner_model_name, 'base_model': base_model_name, 'refiner_model': refiner_model_name,
'denoising_strength': denoising_strength, 'denoising_strength': denoising_strength,
'freeu': advanced_parameters.freeu_enabled, 'freeu': freeu_enabled,
'img2img': input_image_checkbox, 'img2img': input_image_checkbox,
'prompt_expansion': task['expansion'] 'prompt_expansion': task['expansion']
} }
if freeu_enabled:
if advanced_parameters.freeu_enabled:
metadata |= { metadata |= {
'freeu_b1': advanced_parameters.freeu_b1, 'freeu_b2': advanced_parameters.freeu_b2, 'freeu_s1': advanced_parameters.freeu_s1, 'freeu_s2': advanced_parameters.freeu_s2 'freeu_b1': freeu_b1, 'freeu_b2': freeu_b2, 'freeu_s1': freeu_s1, 'freeu_s2': freeu_s2
} }
if 'vary' in goals: if 'vary' in goals:
@@ -851,14 +887,14 @@ def worker():
} }
else: else:
metadata |= { metadata |= {
'inpaint_additional_prompt': inpaint_additional_prompt, 'inpaint_mask_upload': advanced_parameters.inpaint_mask_upload_checkbox, 'invert_mask': advanced_parameters.invert_mask_checkbox, 'inpaint_additional_prompt': inpaint_additional_prompt, 'inpaint_mask_upload': inpaint_mask_upload_checkbox, 'invert_mask': invert_mask_checkbox,
'inpaint_disable_initial_latent': advanced_parameters.inpaint_disable_initial_latent, 'inpaint_engine': advanced_parameters.inpaint_engine, 'inpaint_disable_initial_latent': inpaint_disable_initial_latent, 'inpaint_engine': inpaint_engine,
'inpaint_strength': advanced_parameters.inpaint_strength, 'inpaint_respective_field': advanced_parameters.inpaint_respective_field, 'inpaint_strength': inpaint_strength, 'inpaint_respective_field': inpaint_respective_field,
} }
if 'cn' in goals: if 'cn' in goals:
metadata |= { metadata |= {
'canny_low_threshold': advanced_parameters.canny_low_threshold, 'canny_high_threshold': advanced_parameters.canny_high_threshold, 'canny_low_threshold': canny_low_threshold, 'canny_high_threshold': canny_high_threshold,
} }
ip_list = {x: [] for x in flags.ip_list} ip_list = {x: [] for x in flags.ip_list}
@@ -906,6 +942,11 @@ def worker():
negative_prompt_text = f"\nNegative prompt: {negative_prompt_resolved}" if negative_prompt_resolved else "" negative_prompt_text = f"\nNegative prompt: {negative_prompt_resolved}" if negative_prompt_resolved else ""
metadata_string = f"{positive_prompt_resolved}{negative_prompt_text}\n{generation_params_text}".strip() metadata_string = f"{positive_prompt_resolved}{negative_prompt_text}\n{generation_params_text}".strip()
if modules.config.default_black_out_nsfw or black_out_nsfw:
progressbar_index = int(15.0 + 85.0 * float((current_task_id + 1) * steps) / float(all_steps))
progressbar(async_task, progressbar_index, 'Checking for NSFW content ...')
imgs = censor_batch(imgs)
for x in imgs: for x in imgs:
d = [ d = [
('Prompt', task['log_positive_prompt']), ('Prompt', task['log_positive_prompt']),
@@ -917,15 +958,15 @@ def worker():
('Sharpness', sharpness), ('Sharpness', sharpness),
('Guidance Scale', guidance_scale), ('Guidance Scale', guidance_scale),
('ADM Guidance', str(( ('ADM Guidance', str((
modules.patch.positive_adm_scale, modules.patch.patch_settings[pid].positive_adm_scale,
modules.patch.negative_adm_scale, modules.patch.patch_settings[pid].negative_adm_scale,
modules.patch.adm_scaler_end))), modules.patch.patch_settings[pid].adm_scaler_end))),
('Base Model', base_model_name), ('Base Model', base_model_name),
('Refiner Model', refiner_model_name), ('Refiner Model', refiner_model_name),
('Refiner Switch', refiner_switch), ('Refiner Switch', refiner_switch),
('Sampler', sampler_name), ('Sampler', sampler_name),
('Scheduler', scheduler_name), ('Scheduler', scheduler_name),
('Sampling Steps Override', advanced_parameters.overwrite_step), ('Sampling Steps Override', overwrite_step),
('Seed', task['task_seed']), ('Seed', task['task_seed']),
] ]
for li, (n, w) in enumerate(loras): for li, (n, w) in enumerate(loras):
@@ -934,7 +975,8 @@ def worker():
d.append(('Version', 'v' + fooocus_version.version)) d.append(('Version', 'v' + fooocus_version.version))
img_paths.append(log(x, d, metadata_string, save_metadata_to_images, output_format)) img_paths.append(log(x, d, metadata_string, save_metadata_to_images, output_format))
yield_result(async_task, img_paths, do_not_show_finished_images=len(tasks) == 1, progressbar_index=int(15.0 + 85.0 * float((current_task_id + 1) * steps) / float(all_steps))) yield_result(async_task, img_paths, black_out_nsfw, do_not_show_finished_images=len(tasks) == 1
or disable_intermediate_results or sampler_name == 'lcm')
except ldm_patched.modules.model_management.InterruptProcessingException as e: except ldm_patched.modules.model_management.InterruptProcessingException as e:
if async_task.last_stop == 'skip': if async_task.last_stop == 'skip':
print('User skipped') print('User skipped')
@@ -953,14 +995,20 @@ def worker():
time.sleep(0.01) time.sleep(0.01)
if len(async_tasks) > 0: if len(async_tasks) > 0:
task = async_tasks.pop(0) task = async_tasks.pop(0)
generate_image_grid = task.args.pop(0)
try: try:
handler(task) handler(task)
build_image_wall(task) if generate_image_grid:
build_image_wall(task)
task.yields.append(['finish', task.results]) task.yields.append(['finish', task.results])
pipeline.prepare_text_encoder(async_call=True) pipeline.prepare_text_encoder(async_call=True)
except: except:
traceback.print_exc() traceback.print_exc()
task.yields.append(['finish', task.results]) task.yields.append(['finish', task.results])
finally:
if pid in modules.patch.patch_settings:
del modules.patch.patch_settings[pid]
pass pass
modules/censor.py
-5
View File
@@ -1,8 +1,5 @@
# modified version of https://github.com/AUTOMATIC1111/stable-diffusion-webui-nsfw-censor/blob/master/scripts/censor.py # modified version of https://github.com/AUTOMATIC1111/stable-diffusion-webui-nsfw-censor/blob/master/scripts/censor.py
import numpy as np import numpy as np
import torch
import modules.core as core
from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
from transformers import AutoFeatureExtractor from transformers import AutoFeatureExtractor
@@ -16,8 +13,6 @@ safety_checker = None
def numpy_to_pil(image): def numpy_to_pil(image):
image = (image * 255).round().astype("uint8") image = (image * 255).round().astype("uint8")
#pil_image = Image.fromarray(image, 'RGB')
pil_image = Image.fromarray(image) pil_image = Image.fromarray(image)
return pil_image return pil_image
modules/core.py
+2 -8
View File
@@ -1,8 +1,3 @@
from modules.patch import patch_all
patch_all()
import os import os
import einops import einops
import torch import torch
@@ -16,7 +11,6 @@ import ldm_patched.modules.controlnet
import modules.sample_hijack import modules.sample_hijack
import ldm_patched.modules.samplers import ldm_patched.modules.samplers
import ldm_patched.modules.latent_formats import ldm_patched.modules.latent_formats
import modules.advanced_parameters
from ldm_patched.modules.sd import load_checkpoint_guess_config from ldm_patched.modules.sd import load_checkpoint_guess_config
from ldm_patched.contrib.external import VAEDecode, EmptyLatentImage, VAEEncode, VAEEncodeTiled, VAEDecodeTiled, \ from ldm_patched.contrib.external import VAEDecode, EmptyLatentImage, VAEEncode, VAEEncodeTiled, VAEDecodeTiled, \
@@ -268,7 +262,7 @@ def get_previewer(model):
def ksampler(model, positive, negative, latent, seed=None, steps=30, cfg=7.0, sampler_name='dpmpp_2m_sde_gpu', def ksampler(model, positive, negative, latent, seed=None, steps=30, cfg=7.0, sampler_name='dpmpp_2m_sde_gpu',
scheduler='karras', denoise=1.0, disable_noise=False, start_step=None, last_step=None, scheduler='karras', denoise=1.0, disable_noise=False, start_step=None, last_step=None,
force_full_denoise=False, callback_function=None, refiner=None, refiner_switch=-1, force_full_denoise=False, callback_function=None, refiner=None, refiner_switch=-1,
previewer_start=None, previewer_end=None, sigmas=None, noise_mean=None): previewer_start=None, previewer_end=None, sigmas=None, noise_mean=None, disable_preview=False):
if sigmas is not None: if sigmas is not None:
sigmas = sigmas.clone().to(ldm_patched.modules.model_management.get_torch_device()) sigmas = sigmas.clone().to(ldm_patched.modules.model_management.get_torch_device())
@@ -299,7 +293,7 @@ def ksampler(model, positive, negative, latent, seed=None, steps=30, cfg=7.0, sa
def callback(step, x0, x, total_steps): def callback(step, x0, x, total_steps):
ldm_patched.modules.model_management.throw_exception_if_processing_interrupted() ldm_patched.modules.model_management.throw_exception_if_processing_interrupted()
y = None y = None
if previewer is not None and not modules.advanced_parameters.disable_preview: if previewer is not None and not disable_preview:
y = previewer(x0, previewer_start + step, previewer_end) y = previewer(x0, previewer_start + step, previewer_end)
if callback_function is not None: if callback_function is not None:
callback_function(previewer_start + step, x0, x, previewer_end, y) callback_function(previewer_start + step, x0, x, previewer_end, y)
modules/default_pipeline.py
+11 -6
View File
@@ -315,7 +315,7 @@ def get_candidate_vae(steps, switch, denoise=1.0, refiner_swap_method='joint'):
@torch.no_grad() @torch.no_grad()
@torch.inference_mode() @torch.inference_mode()
def process_diffusion(positive_cond, negative_cond, steps, switch, width, height, image_seed, callback, sampler_name, scheduler_name, latent=None, denoise=1.0, tiled=False, cfg_scale=7.0, refiner_swap_method='joint'): def process_diffusion(positive_cond, negative_cond, steps, switch, width, height, image_seed, callback, sampler_name, scheduler_name, latent=None, denoise=1.0, tiled=False, cfg_scale=7.0, refiner_swap_method='joint', disable_preview=False):
target_unet, target_vae, target_refiner_unet, target_refiner_vae, target_clip \ target_unet, target_vae, target_refiner_unet, target_refiner_vae, target_clip \
= final_unet, final_vae, final_refiner_unet, final_refiner_vae, final_clip = final_unet, final_vae, final_refiner_unet, final_refiner_vae, final_clip
@@ -374,6 +374,7 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
refiner_switch=switch, refiner_switch=switch,
previewer_start=0, previewer_start=0,
previewer_end=steps, previewer_end=steps,
disable_preview=disable_preview
) )
decoded_latent = core.decode_vae(vae=target_vae, latent_image=sampled_latent, tiled=tiled) decoded_latent = core.decode_vae(vae=target_vae, latent_image=sampled_latent, tiled=tiled)
@@ -392,6 +393,7 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
scheduler=scheduler_name, scheduler=scheduler_name,
previewer_start=0, previewer_start=0,
previewer_end=steps, previewer_end=steps,
disable_preview=disable_preview
) )
print('Refiner swapped by changing ksampler. Noise preserved.') print('Refiner swapped by changing ksampler. Noise preserved.')
@@ -414,6 +416,7 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
scheduler=scheduler_name, scheduler=scheduler_name,
previewer_start=switch, previewer_start=switch,
previewer_end=steps, previewer_end=steps,
disable_preview=disable_preview
) )
target_model = target_refiner_vae target_model = target_refiner_vae
@@ -422,7 +425,7 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
decoded_latent = core.decode_vae(vae=target_model, latent_image=sampled_latent, tiled=tiled) decoded_latent = core.decode_vae(vae=target_model, latent_image=sampled_latent, tiled=tiled)
if refiner_swap_method == 'vae': if refiner_swap_method == 'vae':
modules.patch.eps_record = 'vae' modules.patch.patch_settings[os.getpid()].eps_record = 'vae'
if modules.inpaint_worker.current_task is not None: if modules.inpaint_worker.current_task is not None:
modules.inpaint_worker.current_task.unswap() modules.inpaint_worker.current_task.unswap()
@@ -440,7 +443,8 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
sampler_name=sampler_name, sampler_name=sampler_name,
scheduler=scheduler_name, scheduler=scheduler_name,
previewer_start=0, previewer_start=0,
previewer_end=steps previewer_end=steps,
disable_preview=disable_preview
) )
print('Fooocus VAE-based swap.') print('Fooocus VAE-based swap.')
@@ -459,7 +463,7 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
denoise=denoise)[switch:] * k_sigmas denoise=denoise)[switch:] * k_sigmas
len_sigmas = len(sigmas) - 1 len_sigmas = len(sigmas) - 1
noise_mean = torch.mean(modules.patch.eps_record, dim=1, keepdim=True) noise_mean = torch.mean(modules.patch.patch_settings[os.getpid()].eps_record, dim=1, keepdim=True)
if modules.inpaint_worker.current_task is not None: if modules.inpaint_worker.current_task is not None:
modules.inpaint_worker.current_task.swap() modules.inpaint_worker.current_task.swap()
@@ -479,7 +483,8 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
previewer_start=switch, previewer_start=switch,
previewer_end=steps, previewer_end=steps,
sigmas=sigmas, sigmas=sigmas,
noise_mean=noise_mean noise_mean=noise_mean,
disable_preview=disable_preview
) )
target_model = target_refiner_vae target_model = target_refiner_vae
@@ -488,5 +493,5 @@ def process_diffusion(positive_cond, negative_cond, steps, switch, width, height
decoded_latent = core.decode_vae(vae=target_model, latent_image=sampled_latent, tiled=tiled) decoded_latent = core.decode_vae(vae=target_model, latent_image=sampled_latent, tiled=tiled)
images = core.pytorch_to_numpy(decoded_latent) images = core.pytorch_to_numpy(decoded_latent)
modules.patch.eps_record = None modules.patch.patch_settings[os.getpid()].eps_record = None
return images return images
modules/patch.py
+38 -32
View File
@@ -17,7 +17,6 @@ import ldm_patched.controlnet.cldm
import ldm_patched.modules.model_patcher import ldm_patched.modules.model_patcher
import ldm_patched.modules.samplers import ldm_patched.modules.samplers
import ldm_patched.modules.args_parser import ldm_patched.modules.args_parser
import modules.advanced_parameters as advanced_parameters
import warnings import warnings
import safetensors.torch import safetensors.torch
import modules.constants as constants import modules.constants as constants
@@ -29,15 +28,25 @@ from modules.patch_precision import patch_all_precision
from modules.patch_clip import patch_all_clip from modules.patch_clip import patch_all_clip
sharpness = 2.0 class PatchSettings:
def __init__(self,
sharpness=2.0,
adm_scaler_end=0.3,
positive_adm_scale=1.5,
negative_adm_scale=0.8,
controlnet_softness=0.25,
adaptive_cfg=7.0):
self.sharpness = sharpness
self.adm_scaler_end = adm_scaler_end
self.positive_adm_scale = positive_adm_scale
self.negative_adm_scale = negative_adm_scale
self.controlnet_softness = controlnet_softness
self.adaptive_cfg = adaptive_cfg
self.global_diffusion_progress = 0
self.eps_record = None
adm_scaler_end = 0.3
positive_adm_scale = 1.5
negative_adm_scale = 0.8
adaptive_cfg = 7.0 patch_settings = {}
global_diffusion_progress = 0
eps_record = None
def calculate_weight_patched(self, patches, weight, key): def calculate_weight_patched(self, patches, weight, key):
@@ -201,14 +210,13 @@ class BrownianTreeNoiseSamplerPatched:
def compute_cfg(uncond, cond, cfg_scale, t): def compute_cfg(uncond, cond, cfg_scale, t):
global adaptive_cfg pid = os.getpid()
mimic_cfg = float(patch_settings[pid].adaptive_cfg)
mimic_cfg = float(adaptive_cfg)
real_cfg = float(cfg_scale) real_cfg = float(cfg_scale)
real_eps = uncond + real_cfg * (cond - uncond) real_eps = uncond + real_cfg * (cond - uncond)
if cfg_scale > adaptive_cfg: if cfg_scale > patch_settings[pid].adaptive_cfg:
mimicked_eps = uncond + mimic_cfg * (cond - uncond) mimicked_eps = uncond + mimic_cfg * (cond - uncond)
return real_eps * t + mimicked_eps * (1 - t) return real_eps * t + mimicked_eps * (1 - t)
else: else:
@@ -216,13 +224,13 @@ def compute_cfg(uncond, cond, cfg_scale, t):
def patched_sampling_function(model, x, timestep, uncond, cond, cond_scale, model_options=None, seed=None): def patched_sampling_function(model, x, timestep, uncond, cond, cond_scale, model_options=None, seed=None):
global eps_record pid = os.getpid()
if math.isclose(cond_scale, 1.0) and not model_options.get("disable_cfg1_optimization", False): if math.isclose(cond_scale, 1.0) and not model_options.get("disable_cfg1_optimization", False):
final_x0 = calc_cond_uncond_batch(model, cond, None, x, timestep, model_options)[0] final_x0 = calc_cond_uncond_batch(model, cond, None, x, timestep, model_options)[0]
if eps_record is not None: if patch_settings[pid].eps_record is not None:
eps_record = ((x - final_x0) / timestep).cpu() patch_settings[pid].eps_record = ((x - final_x0) / timestep).cpu()
return final_x0 return final_x0
@@ -231,16 +239,16 @@ def patched_sampling_function(model, x, timestep, uncond, cond, cond_scale, mode
positive_eps = x - positive_x0 positive_eps = x - positive_x0
negative_eps = x - negative_x0 negative_eps = x - negative_x0
alpha = 0.001 * sharpness * global_diffusion_progress alpha = 0.001 * patch_settings[pid].sharpness * patch_settings[pid].global_diffusion_progress
positive_eps_degraded = anisotropic.adaptive_anisotropic_filter(x=positive_eps, g=positive_x0) positive_eps_degraded = anisotropic.adaptive_anisotropic_filter(x=positive_eps, g=positive_x0)
positive_eps_degraded_weighted = positive_eps_degraded * alpha + positive_eps * (1.0 - alpha) positive_eps_degraded_weighted = positive_eps_degraded * alpha + positive_eps * (1.0 - alpha)
final_eps = compute_cfg(uncond=negative_eps, cond=positive_eps_degraded_weighted, final_eps = compute_cfg(uncond=negative_eps, cond=positive_eps_degraded_weighted,
cfg_scale=cond_scale, t=global_diffusion_progress) cfg_scale=cond_scale, t=patch_settings[pid].global_diffusion_progress)
if eps_record is not None: if patch_settings[pid].eps_record is not None:
eps_record = (final_eps / timestep).cpu() patch_settings[pid].eps_record = (final_eps / timestep).cpu()
return x - final_eps return x - final_eps
@@ -255,20 +263,19 @@ def round_to_64(x):
def sdxl_encode_adm_patched(self, **kwargs): def sdxl_encode_adm_patched(self, **kwargs):
global positive_adm_scale, negative_adm_scale
clip_pooled = ldm_patched.modules.model_base.sdxl_pooled(kwargs, self.noise_augmentor) clip_pooled = ldm_patched.modules.model_base.sdxl_pooled(kwargs, self.noise_augmentor)
width = kwargs.get("width", 1024) width = kwargs.get("width", 1024)
height = kwargs.get("height", 1024) height = kwargs.get("height", 1024)
target_width = width target_width = width
target_height = height target_height = height
pid = os.getpid()
if kwargs.get("prompt_type", "") == "negative": if kwargs.get("prompt_type", "") == "negative":
width = float(width) * negative_adm_scale width = float(width) * patch_settings[pid].negative_adm_scale
height = float(height) * negative_adm_scale height = float(height) * patch_settings[pid].negative_adm_scale
elif kwargs.get("prompt_type", "") == "positive": elif kwargs.get("prompt_type", "") == "positive":
width = float(width) * positive_adm_scale width = float(width) * patch_settings[pid].positive_adm_scale
height = float(height) * positive_adm_scale height = float(height) * patch_settings[pid].positive_adm_scale
def embedder(number_list): def embedder(number_list):
h = self.embedder(torch.tensor(number_list, dtype=torch.float32)) h = self.embedder(torch.tensor(number_list, dtype=torch.float32))
@@ -322,7 +329,7 @@ def patched_KSamplerX0Inpaint_forward(self, x, sigma, uncond, cond, cond_scale,
def timed_adm(y, timesteps): def timed_adm(y, timesteps):
if isinstance(y, torch.Tensor) and int(y.dim()) == 2 and int(y.shape[1]) == 5632: if isinstance(y, torch.Tensor) and int(y.dim()) == 2 and int(y.shape[1]) == 5632:
y_mask = (timesteps > 999.0 * (1.0 - float(adm_scaler_end))).to(y)[..., None] y_mask = (timesteps > 999.0 * (1.0 - float(patch_settings[os.getpid()].adm_scaler_end))).to(y)[..., None]
y_with_adm = y[..., :2816].clone() y_with_adm = y[..., :2816].clone()
y_without_adm = y[..., 2816:].clone() y_without_adm = y[..., 2816:].clone()
return y_with_adm * y_mask + y_without_adm * (1.0 - y_mask) return y_with_adm * y_mask + y_without_adm * (1.0 - y_mask)
@@ -332,6 +339,7 @@ def timed_adm(y, timesteps):
def patched_cldm_forward(self, x, hint, timesteps, context, y=None, **kwargs): def patched_cldm_forward(self, x, hint, timesteps, context, y=None, **kwargs):
t_emb = ldm_patched.ldm.modules.diffusionmodules.openaimodel.timestep_embedding(timesteps, self.model_channels, repeat_only=False).to(x.dtype) t_emb = ldm_patched.ldm.modules.diffusionmodules.openaimodel.timestep_embedding(timesteps, self.model_channels, repeat_only=False).to(x.dtype)
emb = self.time_embed(t_emb) emb = self.time_embed(t_emb)
pid = os.getpid()
guided_hint = self.input_hint_block(hint, emb, context) guided_hint = self.input_hint_block(hint, emb, context)
@@ -357,19 +365,17 @@ def patched_cldm_forward(self, x, hint, timesteps, context, y=None, **kwargs):
h = self.middle_block(h, emb, context) h = self.middle_block(h, emb, context)
outs.append(self.middle_block_out(h, emb, context)) outs.append(self.middle_block_out(h, emb, context))
if advanced_parameters.controlnet_softness > 0: if patch_settings[pid].controlnet_softness > 0:
for i in range(10): for i in range(10):
k = 1.0 - float(i) / 9.0 k = 1.0 - float(i) / 9.0
outs[i] = outs[i] * (1.0 - advanced_parameters.controlnet_softness * k) outs[i] = outs[i] * (1.0 - patch_settings[pid].controlnet_softness * k)
return outs return outs
def patched_unet_forward(self, x, timesteps=None, context=None, y=None, control=None, transformer_options={}, **kwargs): def patched_unet_forward(self, x, timesteps=None, context=None, y=None, control=None, transformer_options={}, **kwargs):
global global_diffusion_progress
self.current_step = 1.0 - timesteps.to(x) / 999.0 self.current_step = 1.0 - timesteps.to(x) / 999.0
global_diffusion_progress = float(self.current_step.detach().cpu().numpy().tolist()[0]) patch_settings[os.getpid()].global_diffusion_progress = float(self.current_step.detach().cpu().numpy().tolist()[0])
y = timed_adm(y, timesteps) y = timed_adm(y, timesteps)
@@ -483,7 +489,7 @@ def patch_all():
if ldm_patched.modules.model_management.directml_enabled: if ldm_patched.modules.model_management.directml_enabled:
ldm_patched.modules.model_management.lowvram_available = True ldm_patched.modules.model_management.lowvram_available = True
ldm_patched.modules.model_management.OOM_EXCEPTION = Exception ldm_patched.modules.model_management.OOM_EXCEPTION = Exception
patch_all_precision() patch_all_precision()
patch_all_clip() patch_all_clip()
readme.md
+1 -1
View File
@@ -5,7 +5,7 @@ The purpose of this fork is to add new features / fix bugs and contribute back t
Included adjustments: Included adjustments:
* ✨ https://github.com/lllyasviel/Fooocus/pull/958 - NSFW image censoring (config and UI) * ✨ https://github.com/lllyasviel/Fooocus/pull/958 - NSFW image censoring (config and UI)
* 🐛 https://github.com/lllyasviel/Fooocus/pull/981 - prevent users from skipping/stopping other users tasks in queue (multi-user capabilities) * 🐛 https://github.com/lllyasviel/Fooocus/pull/981 - prevent users from skipping/stopping other users tasks in queue (multi-user capabilities) + rework advanced_parameters (removal + PID handling)
* ✨ https://github.com/lllyasviel/Fooocus/pull/985 - add list of 100 animals to wildcards * ✨ https://github.com/lllyasviel/Fooocus/pull/985 - add list of 100 animals to wildcards
* ✨ https://github.com/lllyasviel/Fooocus/pull/1013 - add advanced parameter for disable_intermediate_results (progress_gallery, prevents UI lag when generation is too fast) * ✨ https://github.com/lllyasviel/Fooocus/pull/1013 - add advanced parameter for disable_intermediate_results (progress_gallery, prevents UI lag when generation is too fast)
* ✨ https://github.com/lllyasviel/Fooocus/pull/1039 - add prompt translation * ✨ https://github.com/lllyasviel/Fooocus/pull/1039 - add prompt translation
webui.py
+18 -19
View File
@@ -11,7 +11,6 @@ import modules.async_worker as worker
import modules.constants as constants import modules.constants as constants
import modules.flags as flags import modules.flags as flags
import modules.gradio_hijack as grh import modules.gradio_hijack as grh
import modules.advanced_parameters as advanced_parameters
import modules.style_sorter as style_sorter import modules.style_sorter as style_sorter
import modules.meta_parser import modules.meta_parser
import args_manager import args_manager
@@ -25,9 +24,9 @@ from modules.auth import auth_enabled, check_auth
def get_task(*args): def get_task(*args):
args = list(args) args = list(args)
currentTask = args.pop(0) args.pop(0)
currentTask = worker.AsyncTask(args=args)
return currentTask return worker.AsyncTask(args=args)
def generate_clicked(task): def generate_clicked(task):
import ldm_patched.modules.model_management as model_management import ldm_patched.modules.model_management as model_management
@@ -65,7 +64,7 @@ def generate_clicked(task):
if flag == 'results': if flag == 'results':
yield gr.update(visible=True), \ yield gr.update(visible=True), \
gr.update(visible=True), \ gr.update(visible=True), \
gr.update(visible=True, value=product) if not advanced_parameters.disable_intermediate_results and advanced_parameters.sampler_name != 'lcm' else gr.update(), \ gr.update(visible=True, value=product), \
gr.update(visible=False) gr.update(visible=False)
if flag == 'finish': if flag == 'finish':
yield gr.update(visible=False), \ yield gr.update(visible=False), \
@@ -285,7 +284,7 @@ with shared.gradio_root:
if args_manager.args.disable_image_log: if args_manager.args.disable_image_log:
return gr.update(value='') return gr.update(value='')
return gr.update(value=f'<a href="/file={get_current_html_path()}" target="_blank">\U0001F4DA History Log</a>') return gr.update(value=f'<a href="file={get_current_html_path()}" target="_blank">\U0001F4DA History Log</a>')
history_link = gr.HTML() history_link = gr.HTML()
shared.gradio_root.load(update_history_link,outputs=history_link) shared.gradio_root.load(update_history_link,outputs=history_link)
@@ -484,7 +483,7 @@ with shared.gradio_root:
'(default is 0, always process before any mask invert)') '(default is 0, always process before any mask invert)')
inpaint_mask_upload_checkbox = gr.Checkbox(label='Enable Mask Upload', value=False) inpaint_mask_upload_checkbox = gr.Checkbox(label='Enable Mask Upload', value=False)
invert_mask_checkbox = gr.Checkbox(label='Invert Mask', value=False) invert_mask_checkbox = gr.Checkbox(label='Invert Mask', value=False)
inpaint_ctrls = [debugging_inpaint_preprocessor, inpaint_disable_initial_latent, inpaint_engine, inpaint_ctrls = [debugging_inpaint_preprocessor, inpaint_disable_initial_latent, inpaint_engine,
inpaint_strength, inpaint_respective_field, inpaint_strength, inpaint_respective_field,
inpaint_mask_upload_checkbox, invert_mask_checkbox, inpaint_erode_or_dilate] inpaint_mask_upload_checkbox, invert_mask_checkbox, inpaint_erode_or_dilate]
@@ -501,15 +500,6 @@ with shared.gradio_root:
freeu_s2 = gr.Slider(label='S2', minimum=0, maximum=4, step=0.01, value=0.95) freeu_s2 = gr.Slider(label='S2', minimum=0, maximum=4, step=0.01, value=0.95)
freeu_ctrls = [freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2] freeu_ctrls = [freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2]
adps = [disable_preview, disable_intermediate_results, black_out_nsfw, adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg, sampler_name,
scheduler_name, generate_image_grid, overwrite_step, overwrite_switch, overwrite_width, overwrite_height,
overwrite_vary_strength, overwrite_upscale_strength,
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint,
debugging_cn_preprocessor, skipping_cn_preprocessor, controlnet_softness,
canny_low_threshold, canny_high_threshold, refiner_swap_method]
adps += freeu_ctrls
adps += inpaint_ctrls
def dev_mode_checked(r): def dev_mode_checked(r):
return gr.update(visible=r) return gr.update(visible=r)
@@ -648,8 +638,9 @@ with shared.gradio_root:
inpaint_strength, inpaint_respective_field inpaint_strength, inpaint_respective_field
], show_progress=False, queue=False) ], show_progress=False, queue=False)
ctrls = [ ctrls = [currentTask, generate_image_grid]
currentTask, prompt, negative_prompt, translate_prompts, style_selections, ctrls += [
prompt, negative_prompt, translate_prompts, style_selections,
performance_selection, aspect_ratios_selection, image_number, output_format, image_seed, sharpness, guidance_scale performance_selection, aspect_ratios_selection, image_number, output_format, image_seed, sharpness, guidance_scale
] ]
@@ -657,6 +648,15 @@ with shared.gradio_root:
ctrls += [input_image_checkbox, current_tab] ctrls += [input_image_checkbox, current_tab]
ctrls += [uov_method, uov_input_image] ctrls += [uov_method, uov_input_image]
ctrls += [outpaint_selections, inpaint_input_image, inpaint_additional_prompt, inpaint_mask_image] ctrls += [outpaint_selections, inpaint_input_image, inpaint_additional_prompt, inpaint_mask_image]
ctrls += [disable_preview, disable_intermediate_results, black_out_nsfw]
ctrls += [adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg]
ctrls += [sampler_name, scheduler_name]
ctrls += [overwrite_step, overwrite_switch, overwrite_width, overwrite_height, overwrite_vary_strength]
ctrls += [overwrite_upscale_strength, mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint]
ctrls += [debugging_cn_preprocessor, skipping_cn_preprocessor, canny_low_threshold, canny_high_threshold]
ctrls += [refiner_swap_method, controlnet_softness]
ctrls += freeu_ctrls
ctrls += inpaint_ctrls
if not args_manager.args.disable_metadata: if not args_manager.args.disable_metadata:
ctrls += [save_metadata_to_images, metadata_scheme] ctrls += [save_metadata_to_images, metadata_scheme]
@@ -689,7 +689,6 @@ with shared.gradio_root:
generate_button.click(lambda: (gr.update(visible=True, interactive=True), gr.update(visible=True, interactive=True), gr.update(visible=False, interactive=False), [], True), generate_button.click(lambda: (gr.update(visible=True, interactive=True), gr.update(visible=True, interactive=True), gr.update(visible=False, interactive=False), [], True),
outputs=[stop_button, skip_button, generate_button, gallery, state_is_generating]) \ outputs=[stop_button, skip_button, generate_button, gallery, state_is_generating]) \
.then(fn=refresh_seed, inputs=[seed_random, image_seed], outputs=image_seed) \ .then(fn=refresh_seed, inputs=[seed_random, image_seed], outputs=image_seed) \
.then(advanced_parameters.set_all_advanced_parameters, inputs=adps) \
.then(fn=get_task, inputs=ctrls, outputs=currentTask) \ .then(fn=get_task, inputs=ctrls, outputs=currentTask) \
.then(fn=generate_clicked, inputs=currentTask, outputs=[progress_html, progress_window, progress_gallery, gallery]) \ .then(fn=generate_clicked, inputs=currentTask, outputs=[progress_html, progress_window, progress_gallery, gallery]) \
.then(lambda: (gr.update(visible=True, interactive=True), gr.update(visible=False, interactive=False), gr.update(visible=False, interactive=False), False), .then(lambda: (gr.update(visible=True, interactive=True), gr.update(visible=False, interactive=False), gr.update(visible=False, interactive=False), False),