rework refiner for some potential new features (#642)
* sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync * sync
This commit is contained in:
lllyasviel
GitHub
parent
777510dc9b
commit
bbdf4bd120
|
|
@ -107,6 +107,10 @@ class ModelPatcher:
|
|||
for k in patch_list:
|
||||
if hasattr(patch_list[k], "to"):
|
||||
patch_list[k] = patch_list[k].to(device)
|
||||
if "unet_wrapper_function" in self.model_options:
|
||||
wrap_func = self.model_options["unet_wrapper_function"]
|
||||
if hasattr(wrap_func, "to"):
|
||||
self.model_options["unet_wrapper_function"] = wrap_func.to(device)
|
||||
|
||||
def model_dtype(self):
|
||||
if hasattr(self.model, "get_dtype"):
|
||||
|
|
|
|||
|
|
@ -22,7 +22,7 @@ def get_empty_folder(path):
|
|||
|
||||
|
||||
comfy_repo = "https://github.com/comfyanonymous/ComfyUI"
|
||||
comfy_commit_hash = "5e885bd9c822a3cc7d50d75a40958265f497cd03"
|
||||
comfy_commit_hash = "8cc75c64ff7188ce72cd4ba595119586e425c09f"
|
||||
comfy_temp_path = get_empty_folder(os.path.join(os.path.dirname(os.path.abspath(__file__)), 'backend', 'comfy_temp'))
|
||||
comfy_core_path = get_empty_folder(os.path.join(os.path.dirname(os.path.abspath(__file__)), 'backend', 'headless'))
|
||||
|
||||
|
|
|
|||
|
|
@ -1 +1 @@
|
|||
version = '2.1.47'
|
||||
version = '2.1.48'
|
||||
|
|
|
|||
|
|
@ -3,7 +3,8 @@ adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg, sampler_
|
|||
overwrite_vary_strength, overwrite_upscale_strength, \
|
||||
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint, \
|
||||
debugging_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold, inpaint_engine, \
|
||||
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2 = [None] * 24
|
||||
refiner_swap_method, \
|
||||
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2 = [None] * 25
|
||||
|
||||
|
||||
def set_all_advanced_parameters(*args):
|
||||
|
|
@ -12,6 +13,7 @@ def set_all_advanced_parameters(*args):
|
|||
overwrite_vary_strength, overwrite_upscale_strength, \
|
||||
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint, \
|
||||
debugging_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold, inpaint_engine, \
|
||||
refiner_swap_method, \
|
||||
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2
|
||||
|
||||
adm_scaler_positive, adm_scaler_negative, adm_scaler_end, adaptive_cfg, sampler_name, \
|
||||
|
|
@ -19,6 +21,7 @@ def set_all_advanced_parameters(*args):
|
|||
overwrite_vary_strength, overwrite_upscale_strength, \
|
||||
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint, \
|
||||
debugging_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold, inpaint_engine, \
|
||||
refiner_swap_method, \
|
||||
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2 = args
|
||||
|
||||
return
|
||||
|
|
|
|||
|
|
@ -513,7 +513,8 @@ def worker():
|
|||
latent=initial_latent,
|
||||
denoise=denoising_strength,
|
||||
tiled=tiled,
|
||||
cfg_scale=cfg_scale
|
||||
cfg_scale=cfg_scale,
|
||||
refiner_swap_method=advanced_parameters.refiner_swap_method
|
||||
)
|
||||
|
||||
del task['c'], task['uc'], positive_cond, negative_cond # Save memory
|
||||
|
|
|
|||
|
|
@ -4,7 +4,6 @@ patch_all()
|
|||
|
||||
|
||||
import os
|
||||
import random
|
||||
import einops
|
||||
import torch
|
||||
import numpy as np
|
||||
|
|
@ -63,47 +62,6 @@ def apply_controlnet(positive, negative, control_net, image, strength, start_per
|
|||
image=image, strength=strength, start_percent=start_percent, end_percent=end_percent)
|
||||
|
||||
|
||||
@torch.no_grad()
|
||||
@torch.inference_mode()
|
||||
def load_unet_only(unet_path):
|
||||
sd_raw = comfy.utils.load_torch_file(unet_path)
|
||||
sd = {}
|
||||
flag = 'model.diffusion_model.'
|
||||
for k in list(sd_raw.keys()):
|
||||
if k.startswith(flag):
|
||||
sd[k[len(flag):]] = sd_raw[k]
|
||||
del sd_raw[k]
|
||||
|
||||
parameters = comfy.utils.calculate_parameters(sd)
|
||||
fp16 = comfy.model_management.should_use_fp16(model_params=parameters)
|
||||
if "input_blocks.0.0.weight" in sd:
|
||||
# ldm
|
||||
model_config = comfy.model_detection.model_config_from_unet(sd, "", fp16)
|
||||
if model_config is None:
|
||||
raise RuntimeError("ERROR: Could not detect model type of: {}".format(unet_path))
|
||||
new_sd = sd
|
||||
else:
|
||||
# diffusers
|
||||
model_config = comfy.model_detection.model_config_from_diffusers_unet(sd, fp16)
|
||||
if model_config is None:
|
||||
print("ERROR UNSUPPORTED UNET", unet_path)
|
||||
return None
|
||||
|
||||
diffusers_keys = comfy.utils.unet_to_diffusers(model_config.unet_config)
|
||||
|
||||
new_sd = {}
|
||||
for k in diffusers_keys:
|
||||
if k in sd:
|
||||
new_sd[diffusers_keys[k]] = sd.pop(k)
|
||||
else:
|
||||
print(diffusers_keys[k], k)
|
||||
offload_device = comfy.model_management.unet_offload_device()
|
||||
model = model_config.get_model(new_sd, "")
|
||||
model = model.to(offload_device)
|
||||
model.load_model_weights(new_sd, "")
|
||||
return comfy.model_patcher.ModelPatcher(model, load_device=comfy.model_management.get_torch_device(), offload_device=offload_device)
|
||||
|
||||
|
||||
@torch.no_grad()
|
||||
@torch.inference_mode()
|
||||
def load_model(ckpt_filename):
|
||||
|
|
@ -239,7 +197,7 @@ def get_previewer():
|
|||
@torch.inference_mode()
|
||||
def ksampler(model, positive, negative, latent, seed=None, steps=30, cfg=7.0, sampler_name='dpmpp_fooocus_2m_sde_inpaint_seamless',
|
||||
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):
|
||||
latent_image = latent["samples"]
|
||||
if disable_noise:
|
||||
noise = torch.zeros(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
|
||||
|
|
@ -253,13 +211,19 @@ def ksampler(model, positive, negative, latent, seed=None, steps=30, cfg=7.0, sa
|
|||
|
||||
previewer = get_previewer()
|
||||
|
||||
if previewer_start is None:
|
||||
previewer_start = 0
|
||||
|
||||
if previewer_end is None:
|
||||
previewer_end = steps
|
||||
|
||||
def callback(step, x0, x, total_steps):
|
||||
comfy.model_management.throw_exception_if_processing_interrupted()
|
||||
y = None
|
||||
if previewer is not None:
|
||||
y = previewer(x0, step, total_steps)
|
||||
y = previewer(x0, previewer_start + step, previewer_end)
|
||||
if callback_function is not None:
|
||||
callback_function(step, x0, x, total_steps, y)
|
||||
callback_function(previewer_start + step, x0, x, previewer_end, y)
|
||||
|
||||
disable_pbar = False
|
||||
modules.sample_hijack.current_refiner = refiner
|
||||
|
|
|
|||
|
|
@ -4,9 +4,9 @@ import torch
|
|||
import modules.path
|
||||
import comfy.model_management
|
||||
|
||||
from comfy.model_patcher import ModelPatcher
|
||||
from comfy.model_base import SDXL, SDXLRefiner
|
||||
from modules.expansion import FooocusExpansion
|
||||
from modules.sample_hijack import clip_separate
|
||||
|
||||
|
||||
xl_base: core.StableDiffusionModel = None
|
||||
|
|
@ -15,14 +15,15 @@ xl_base_hash = ''
|
|||
xl_base_patched: core.StableDiffusionModel = None
|
||||
xl_base_patched_hash = ''
|
||||
|
||||
xl_refiner: ModelPatcher = None
|
||||
xl_refiner: core.StableDiffusionModel = None
|
||||
xl_refiner_hash = ''
|
||||
|
||||
final_expansion = None
|
||||
final_unet = None
|
||||
final_clip = None
|
||||
final_vae = None
|
||||
final_refiner = None
|
||||
final_refiner_unet = None
|
||||
final_refiner_vae = None
|
||||
|
||||
loaded_ControlNets = {}
|
||||
|
||||
|
|
@ -60,8 +61,10 @@ def assert_model_integrity():
|
|||
error_message = 'You have selected base model other than SDXL. This is not supported yet.'
|
||||
|
||||
if xl_refiner is not None:
|
||||
if not isinstance(xl_refiner.model, SDXLRefiner):
|
||||
error_message = 'You have selected refiner model other than SDXL refiner. This is not supported yet.'
|
||||
if xl_refiner.unet is None or xl_refiner.unet.model is None:
|
||||
error_message = 'You have selected an invalid refiner!'
|
||||
elif not isinstance(xl_refiner.unet.model, SDXL) and not isinstance(xl_refiner.unet.model, SDXLRefiner):
|
||||
error_message = 'SD1.5 or 2.1 as refiner is not supported!'
|
||||
|
||||
if error_message is not None:
|
||||
raise NotImplementedError(error_message)
|
||||
|
|
@ -109,9 +112,19 @@ def refresh_refiner_model(name):
|
|||
print(f'Refiner unloaded.')
|
||||
return
|
||||
|
||||
xl_refiner = core.load_unet_only(filename)
|
||||
xl_refiner = core.load_model(filename)
|
||||
xl_refiner_hash = model_hash
|
||||
print(f'Refiner model loaded: {model_hash}')
|
||||
|
||||
if isinstance(xl_refiner.unet.model, SDXL):
|
||||
xl_refiner.clip = None
|
||||
xl_refiner.vae = None
|
||||
elif isinstance(xl_refiner.unet.model, SDXLRefiner):
|
||||
xl_refiner.clip = None
|
||||
xl_refiner.vae = None
|
||||
else:
|
||||
xl_refiner = None # 1.5/2.1 not supported yet.
|
||||
|
||||
return
|
||||
|
||||
|
||||
|
|
@ -203,15 +216,23 @@ def prepare_text_encoder(async_call=True):
|
|||
@torch.no_grad()
|
||||
@torch.inference_mode()
|
||||
def refresh_everything(refiner_model_name, base_model_name, loras):
|
||||
global final_unet, final_clip, final_vae, final_refiner, final_expansion
|
||||
global final_unet, final_clip, final_vae, final_refiner_unet, final_refiner_vae, final_expansion
|
||||
|
||||
refresh_refiner_model(refiner_model_name)
|
||||
refresh_base_model(base_model_name)
|
||||
refresh_loras(loras)
|
||||
assert_model_integrity()
|
||||
|
||||
final_unet, final_clip, final_vae, final_refiner = \
|
||||
xl_base_patched.unet, xl_base_patched.clip, xl_base_patched.vae, xl_refiner
|
||||
final_unet = xl_base_patched.unet
|
||||
final_clip = xl_base_patched.clip
|
||||
final_vae = xl_base_patched.vae
|
||||
|
||||
if xl_refiner is None:
|
||||
final_refiner_unet = None
|
||||
final_refiner_vae = None
|
||||
else:
|
||||
final_refiner_unet = xl_refiner.unet
|
||||
final_refiner_vae = xl_refiner.vae
|
||||
|
||||
if final_expansion is None:
|
||||
final_expansion = FooocusExpansion()
|
||||
|
|
@ -236,30 +257,141 @@ refresh_everything(
|
|||
|
||||
@torch.no_grad()
|
||||
@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):
|
||||
def vae_parse(x, tiled=False):
|
||||
if final_vae is None or final_refiner_vae is None:
|
||||
return x
|
||||
|
||||
print('VAE parsing ...')
|
||||
x = core.decode_vae(vae=final_vae, latent_image=x, tiled=tiled)
|
||||
x = core.encode_vae(vae=final_refiner_vae, pixels=x, tiled=tiled)
|
||||
print('VAE parsed ...')
|
||||
|
||||
return x
|
||||
|
||||
|
||||
@torch.no_grad()
|
||||
@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'):
|
||||
assert refiner_swap_method in ['joint', 'separate', 'vae']
|
||||
print(f'[Sampler] refiner_swap_method = {refiner_swap_method}')
|
||||
|
||||
if latent is None:
|
||||
empty_latent = core.generate_empty_latent(width=width, height=height, batch_size=1)
|
||||
else:
|
||||
empty_latent = latent
|
||||
|
||||
sampled_latent = core.ksampler(
|
||||
model=final_unet,
|
||||
refiner=final_refiner,
|
||||
positive=positive_cond,
|
||||
negative=negative_cond,
|
||||
latent=empty_latent,
|
||||
steps=steps, start_step=0, last_step=steps, disable_noise=False, force_full_denoise=True,
|
||||
seed=image_seed,
|
||||
denoise=denoise,
|
||||
callback_function=callback,
|
||||
cfg=cfg_scale,
|
||||
sampler_name=sampler_name,
|
||||
scheduler=scheduler_name,
|
||||
refiner_switch=switch
|
||||
)
|
||||
if refiner_swap_method == 'joint':
|
||||
sampled_latent = core.ksampler(
|
||||
model=final_unet,
|
||||
refiner=final_refiner_unet,
|
||||
positive=positive_cond,
|
||||
negative=negative_cond,
|
||||
latent=empty_latent,
|
||||
steps=steps, start_step=0, last_step=steps, disable_noise=False, force_full_denoise=True,
|
||||
seed=image_seed,
|
||||
denoise=denoise,
|
||||
callback_function=callback,
|
||||
cfg=cfg_scale,
|
||||
sampler_name=sampler_name,
|
||||
scheduler=scheduler_name,
|
||||
refiner_switch=switch,
|
||||
previewer_start=0,
|
||||
previewer_end=steps,
|
||||
)
|
||||
decoded_latent = core.decode_vae(vae=final_vae, latent_image=sampled_latent, tiled=tiled)
|
||||
images = core.pytorch_to_numpy(decoded_latent)
|
||||
return images
|
||||
|
||||
decoded_latent = core.decode_vae(vae=final_vae, latent_image=sampled_latent, tiled=tiled)
|
||||
images = core.pytorch_to_numpy(decoded_latent)
|
||||
if refiner_swap_method == 'separate':
|
||||
sampled_latent = core.ksampler(
|
||||
model=final_unet,
|
||||
positive=positive_cond,
|
||||
negative=negative_cond,
|
||||
latent=empty_latent,
|
||||
steps=steps, start_step=0, last_step=switch, disable_noise=False, force_full_denoise=False,
|
||||
seed=image_seed,
|
||||
denoise=denoise,
|
||||
callback_function=callback,
|
||||
cfg=cfg_scale,
|
||||
sampler_name=sampler_name,
|
||||
scheduler=scheduler_name,
|
||||
previewer_start=0,
|
||||
previewer_end=switch,
|
||||
)
|
||||
print('Refiner swapped by changing ksampler. Noise preserved.')
|
||||
|
||||
comfy.model_management.soft_empty_cache()
|
||||
return images
|
||||
target_model = final_refiner_unet
|
||||
if target_model is None:
|
||||
target_model = final_unet
|
||||
print('Use base model to refine itself - this may because of developer mode.')
|
||||
|
||||
sampled_latent = core.ksampler(
|
||||
model=target_model,
|
||||
positive=clip_separate(positive_cond, target_model=target_model.model),
|
||||
negative=clip_separate(negative_cond, target_model=target_model.model),
|
||||
latent=sampled_latent,
|
||||
steps=steps, start_step=switch, last_step=steps, disable_noise=True, force_full_denoise=True,
|
||||
seed=image_seed,
|
||||
denoise=denoise,
|
||||
callback_function=callback,
|
||||
cfg=cfg_scale,
|
||||
sampler_name=sampler_name,
|
||||
scheduler=scheduler_name,
|
||||
previewer_start=switch,
|
||||
previewer_end=steps,
|
||||
)
|
||||
|
||||
target_model = final_refiner_vae
|
||||
if target_model is None:
|
||||
target_model = final_vae
|
||||
decoded_latent = core.decode_vae(vae=target_model, latent_image=sampled_latent, tiled=tiled)
|
||||
images = core.pytorch_to_numpy(decoded_latent)
|
||||
return images
|
||||
|
||||
if refiner_swap_method == 'vae':
|
||||
sampled_latent = core.ksampler(
|
||||
model=final_unet,
|
||||
positive=positive_cond,
|
||||
negative=negative_cond,
|
||||
latent=empty_latent,
|
||||
steps=steps, start_step=0, last_step=switch, disable_noise=False, force_full_denoise=True,
|
||||
seed=image_seed,
|
||||
denoise=denoise,
|
||||
callback_function=callback,
|
||||
cfg=cfg_scale,
|
||||
sampler_name=sampler_name,
|
||||
scheduler=scheduler_name,
|
||||
previewer_start=0,
|
||||
previewer_end=switch,
|
||||
)
|
||||
print('Refiner swapped by changing ksampler. Noise is not preserved.')
|
||||
|
||||
target_model = final_refiner_unet
|
||||
if target_model is None:
|
||||
target_model = final_unet
|
||||
print('Use base model to refine itself - this may because of developer mode.')
|
||||
|
||||
sampled_latent = vae_parse(sampled_latent)
|
||||
|
||||
sampled_latent = core.ksampler(
|
||||
model=target_model,
|
||||
positive=clip_separate(positive_cond, target_model=target_model.model),
|
||||
negative=clip_separate(negative_cond, target_model=target_model.model),
|
||||
latent=sampled_latent,
|
||||
steps=steps, start_step=switch, last_step=steps, disable_noise=False, force_full_denoise=True,
|
||||
seed=image_seed,
|
||||
denoise=denoise,
|
||||
callback_function=callback,
|
||||
cfg=cfg_scale,
|
||||
sampler_name=sampler_name,
|
||||
scheduler=scheduler_name,
|
||||
previewer_start=switch,
|
||||
previewer_end=steps,
|
||||
)
|
||||
|
||||
target_model = final_refiner_vae
|
||||
if target_model is None:
|
||||
target_model = final_vae
|
||||
decoded_latent = core.decode_vae(vae=target_model, latent_image=sampled_latent, tiled=tiled)
|
||||
images = core.pytorch_to_numpy(decoded_latent)
|
||||
return images
|
||||
|
|
|
|||
|
|
@ -83,16 +83,13 @@ def area_abcd(a, b, c, d):
|
|||
return (b - a) * (d - c)
|
||||
|
||||
|
||||
def solve_abcd(x, a, b, c, d, k, outpaint):
|
||||
def solve_abcd(x, a, b, c, d, outpaint):
|
||||
H, W = x.shape[:2]
|
||||
if outpaint:
|
||||
return 0, H, 0, W
|
||||
min_area = H * W * k
|
||||
max_area = H * W
|
||||
min_area = (min(H, W) ** 2) * 0.5
|
||||
while True:
|
||||
if area_abcd(a, b, c, d) > min_area and abs((b - a) - (d - c)) < 16:
|
||||
break
|
||||
if area_abcd(a, b, c, d) >= max_area:
|
||||
if area_abcd(a, b, c, d) >= min_area:
|
||||
break
|
||||
|
||||
add_h = (b - a) < (d - c)
|
||||
|
|
@ -150,7 +147,7 @@ class InpaintWorker:
|
|||
|
||||
# compute abcd
|
||||
a, b, c, d = compute_initial_abcd(self.mask_raw_bg < 127)
|
||||
a, b, c, d = solve_abcd(self.mask_raw_bg, a, b, c, d, k=0.618, outpaint=is_outpaint)
|
||||
a, b, c, d = solve_abcd(self.mask_raw_bg, a, b, c, d, outpaint=is_outpaint)
|
||||
|
||||
# interested area
|
||||
self.interested_area = (a, b, c, d)
|
||||
|
|
|
|||
|
|
@ -2,7 +2,8 @@ import torch
|
|||
import comfy.samplers
|
||||
import comfy.model_management
|
||||
|
||||
from comfy.sample import prepare_sampling, cleanup_additional_models, get_additional_models
|
||||
from comfy.model_base import SDXLRefiner, SDXL
|
||||
from comfy.sample import get_additional_models
|
||||
from comfy.samplers import resolve_areas_and_cond_masks, wrap_model, calculate_start_end_timesteps, \
|
||||
create_cond_with_same_area_if_none, pre_run_control, apply_empty_x_to_equal_area, encode_adm, \
|
||||
blank_inpaint_image_like
|
||||
|
|
@ -14,11 +15,18 @@ refiner_switch_step = -1
|
|||
|
||||
@torch.no_grad()
|
||||
@torch.inference_mode()
|
||||
def clip_separate(cond):
|
||||
def clip_separate(cond, target_model=None):
|
||||
c, p = cond[0]
|
||||
c = c[..., -1280:].clone()
|
||||
p = p["pooled_output"].clone()
|
||||
return [[c, {"pooled_output": p}]]
|
||||
if target_model is None or isinstance(target_model, SDXLRefiner):
|
||||
c = c[..., -1280:].clone()
|
||||
p = {"pooled_output": p["pooled_output"].clone()}
|
||||
elif isinstance(target_model, SDXL):
|
||||
c = c.clone()
|
||||
p = {"pooled_output": p["pooled_output"].clone()}
|
||||
else:
|
||||
c = c[..., :768].clone()
|
||||
p = {}
|
||||
return [[c, p]]
|
||||
|
||||
|
||||
@torch.no_grad()
|
||||
|
|
@ -54,8 +62,11 @@ def sample_hacked(model, noise, positive, negative, cfg, device, sampler, sigmas
|
|||
negative = encode_adm(model, negative, noise.shape[0], noise.shape[3], noise.shape[2], device, "negative")
|
||||
|
||||
if current_refiner is not None and current_refiner.model.is_adm():
|
||||
positive_refiner = encode_adm(current_refiner.model, clip_separate(positive), noise.shape[0], noise.shape[3], noise.shape[2], device, "positive")
|
||||
negative_refiner = encode_adm(current_refiner.model, clip_separate(negative), noise.shape[0], noise.shape[3], noise.shape[2], device, "negative")
|
||||
positive_refiner = clip_separate(positive, target_model=current_refiner.model)
|
||||
negative_refiner = clip_separate(negative, target_model=current_refiner.model)
|
||||
|
||||
positive_refiner = encode_adm(current_refiner.model, positive_refiner, noise.shape[0], noise.shape[3], noise.shape[2], device, "positive")
|
||||
negative_refiner = encode_adm(current_refiner.model, negative_refiner, noise.shape[0], noise.shape[3], noise.shape[2], device, "negative")
|
||||
|
||||
positive_refiner[0][1]['adm_encoded'].to(positive[0][1]['adm_encoded'])
|
||||
negative_refiner[0][1]['adm_encoded'].to(negative[0][1]['adm_encoded'])
|
||||
|
|
|
|||
6
webui.py
6
webui.py
|
|
@ -230,6 +230,10 @@ with shared.gradio_root:
|
|||
adm_scaler_end = gr.Slider(label='ADM Guidance End At Step', minimum=0.0, maximum=1.0,
|
||||
step=0.001, value=0.3,
|
||||
info='When to end the guidance from positive/negative ADM. ')
|
||||
|
||||
refiner_swap_method = gr.Dropdown(label='Refiner swap method', value='joint',
|
||||
choices=['joint', 'separate', 'vae'])
|
||||
|
||||
adaptive_cfg = gr.Slider(label='CFG Mimicking from TSNR', minimum=1.0, maximum=30.0, step=0.01, value=7.0,
|
||||
info='Enabling Fooocus\'s implementation of CFG mimicking for TSNR '
|
||||
'(effective when real CFG > mimicked CFG).')
|
||||
|
|
@ -293,7 +297,7 @@ with shared.gradio_root:
|
|||
overwrite_vary_strength, overwrite_upscale_strength,
|
||||
mixing_image_prompt_and_vary_upscale, mixing_image_prompt_and_inpaint,
|
||||
debugging_cn_preprocessor, controlnet_softness, canny_low_threshold, canny_high_threshold,
|
||||
inpaint_engine]
|
||||
inpaint_engine, refiner_swap_method]
|
||||
adps += freeu_ctrls
|
||||
|
||||
def dev_mode_checked(r):
|
||||
|
|
|
|||
Loading…
Reference in New Issue