Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 01:04:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

add remote vae

Browse Source 
Signed-off-by: Vladimir Mandic <mandic00@live.com>
This commit is contained in:
Vladimir Mandic
2025-02-22 12:50:18 -05:00
parent f8f987fed6
commit 1b2d4286b5
21 changed files with 133 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/control/run.py
+2 -2
View File
@@ -228,7 +228,7 @@ def control_run(state: str = '',
steps: int = 20, sampler_index: int = None,
seed: int = -1, subseed: int = -1, subseed_strength: float = 0, seed_resize_from_h: int = -1, seed_resize_from_w: int = -1,
cfg_scale: float = 6.0, clip_skip: float = 1.0, image_cfg_scale: float = 6.0, diffusers_guidance_rescale: float = 0.7, pag_scale: float = 0.0, pag_adaptive: float = 0.5, cfg_end: float = 1.0,
full_quality: bool = True, tiling: bool = False, hidiffusion: bool = False,
vae_type: str = 'Full', tiling: bool = False, hidiffusion: bool = False,
detailer_enabled: bool = True, detailer_prompt: str = '', detailer_negative: str = '', detailer_steps: int = 10, detailer_strength: float = 0.3,
hdr_mode: int = 0, hdr_brightness: float = 0, hdr_color: float = 0, hdr_sharpen: float = 0, hdr_clamp: bool = False, hdr_boundary: float = 4.0, hdr_threshold: float = 0.95,
hdr_maximize: bool = False, hdr_max_center: float = 0.6, hdr_max_boundry: float = 1.0, hdr_color_picker: str = None, hdr_tint_ratio: float = 0,
@@ -292,7 +292,7 @@ def control_run(state: str = '',
diffusers_guidance_rescale = diffusers_guidance_rescale,
pag_scale = pag_scale,
pag_adaptive = pag_adaptive,
full_quality = full_quality,
vae_type = vae_type,
tiling = tiling,
hidiffusion = hidiffusion,
# detailer
modules/images_resize.py
+2 -2
View File
@@ -16,9 +16,9 @@ def resize_image(resize_mode: int, im: Union[Image.Image, torch.Tensor], width:
return im
else:
from modules.processing_vae import vae_encode, vae_decode
latents = vae_encode(im, shared.sd_model, full_quality=False) # TODO resize image: enable full VAE mode for resize-latent
latents = vae_encode(im, shared.sd_model, vae_type='Tiny') # TODO resize image: enable full VAE mode for resize-latent
latents = selected_upscaler.scaler.upscale(latents, scale, selected_upscaler.name)
im = vae_decode(latents, shared.sd_model, output_type='pil', full_quality=False)[0]
im = vae_decode(latents, shared.sd_model, output_type='pil', vae_type='Tiny')[0]
return im
def resize(im: Union[Image.Image, torch.Tensor], w, h):
modules/img2img.py
+2 -2
View File
@@ -139,7 +139,7 @@ def img2img(id_task: str, state: str, mode: int,
sampler_index,
mask_blur, mask_alpha,
inpainting_fill,
full_quality, tiling, hidiffusion,
vae_type, tiling, hidiffusion,
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength,
n_iter, batch_size,
cfg_scale, image_cfg_scale,
@@ -241,7 +241,7 @@ def img2img(id_task: str, state: str, mode: int,
clip_skip=clip_skip,
width=width,
height=height,
full_quality=full_quality,
vae_type=vae_type,
tiling=tiling,
hidiffusion=hidiffusion,
detailer_enabled=detailer_enabled,
modules/infotext.py
+1 -1
View File
@@ -105,7 +105,7 @@ def parse(infotext):
elif val == "False":
params[key] = False
elif key == 'VAE' and val == 'TAESD':
params["Full quality"] = False
params["VAE type"] = 'Tiny'
elif size is not None:
params[f"{key}-1"] = int(size.group(1))
params[f"{key}-2"] = int(size.group(2))
modules/processing.py
+1 -1
View File
@@ -151,7 +151,7 @@ def process_images(p: StableDiffusionProcessing) -> Processed:
sd_models.reload_model_weights()
if p.override_settings.get('sd_vae', None) is not None:
if p.override_settings.get('sd_vae', None) == 'TAESD':
p.full_quality = False
p.vae_type = 'Tiny'
p.override_settings.pop('sd_vae', None)
if p.override_settings.get('Hires upscaler', None) is not None:
p.enable_hr = True
modules/processing_args.py
+1 -1
View File
@@ -79,7 +79,7 @@ def task_specific_kwargs(p, model):
}
if model.__class__.__name__ == 'LatentConsistencyModelPipeline' and hasattr(p, 'init_images') and len(p.init_images) > 0:
p.ops.append('lcm')
init_latents = [processing_vae.vae_encode(image, model=shared.sd_model, full_quality=p.full_quality).squeeze(dim=0) for image in p.init_images]
init_latents = [processing_vae.vae_encode(image, model=shared.sd_model, vae_type=p.vae_type).squeeze(dim=0) for image in p.init_images]
init_latent = torch.stack(init_latents, dim=0).to(shared.device)
init_noise = p.denoising_strength * processing.create_random_tensors(init_latent.shape[1:], seeds=p.all_seeds, subseeds=p.all_subseeds, subseed_strength=p.subseed_strength, p=p)
init_latent = (1 - p.denoising_strength) * init_latent + init_noise
modules/processing_class.py
+2 -2
View File
@@ -48,7 +48,7 @@ class StableDiffusionProcessing:
styles: List[str] = [],
# vae
tiling: bool = False,
full_quality: bool = True,
vae_type: str = 'Full',
# other
hidiffusion: bool = False,
do_not_reload_embeddings: bool = False,
@@ -169,7 +169,7 @@ class StableDiffusionProcessing:
self.negative_prompt = negative_prompt
self.styles = styles
self.tiling = tiling
self.full_quality = full_quality
self.vae_type = vae_type
self.hidiffusion = hidiffusion
self.do_not_reload_embeddings = do_not_reload_embeddings
self.detailer_enabled = detailer_enabled
modules/processing_diffusers.py
+5 -5
View File
@@ -197,10 +197,10 @@ def process_hires(p: processing.StableDiffusionProcessing, output):
if p.hr_force:
shared.sd_model = sd_models.set_diffuser_pipe(shared.sd_model, sd_models.DiffusersTaskType.IMAGE_2_IMAGE)
if 'Upscale' in shared.sd_model.__class__.__name__ or 'Flux' in shared.sd_model.__class__.__name__ or 'Kandinsky' in shared.sd_model.__class__.__name__:
output.images = processing_vae.vae_decode(latents=output.images, model=shared.sd_model, full_quality=p.full_quality, output_type='pil', width=p.width, height=p.height)
output.images = processing_vae.vae_decode(latents=output.images, model=shared.sd_model, vae_type=p.vae_type, output_type='pil', width=p.width, height=p.height)
if p.is_control and hasattr(p, 'task_args') and p.task_args.get('image', None) is not None:
if hasattr(shared.sd_model, "vae") and output.images is not None and len(output.images) > 0:
output.images = processing_vae.vae_decode(latents=output.images, model=shared.sd_model, full_quality=p.full_quality, output_type='pil', width=p.hr_upscale_to_x, height=p.hr_upscale_to_y) # controlnet cannnot deal with latent input
output.images = processing_vae.vae_decode(latents=output.images, model=shared.sd_model, vae_type=p.vae_type, output_type='pil', width=p.hr_upscale_to_x, height=p.hr_upscale_to_y) # controlnet cannnot deal with latent input
update_sampler(p, shared.sd_model, second_pass=True)
orig_denoise = p.denoising_strength
p.denoising_strength = strength
@@ -289,7 +289,7 @@ def process_refine(p: processing.StableDiffusionProcessing, output):
noise_level = round(350 * p.denoising_strength)
output_type='latent'
if 'Upscale' in shared.sd_refiner.__class__.__name__ or 'Flux' in shared.sd_refiner.__class__.__name__ or 'Kandinsky' in shared.sd_refiner.__class__.__name__:
image = processing_vae.vae_decode(latents=image, model=shared.sd_model, full_quality=p.full_quality, output_type='pil', width=p.width, height=p.height)
image = processing_vae.vae_decode(latents=image, model=shared.sd_model, vae_type=p.vae_type, output_type='pil', width=p.width, height=p.height)
p.extra_generation_params['Noise level'] = noise_level
output_type = 'np'
update_sampler(p, shared.sd_refiner, second_pass=True)
@@ -370,7 +370,7 @@ def process_decode(p: processing.StableDiffusionProcessing, output):
result_batch = processing_vae.vae_decode(
latents = output.images[i],
model = model,
full_quality = p.full_quality,
vae_type = p.vae_type,
width = width,
height = height,
frames = frames,
@@ -381,7 +381,7 @@ def process_decode(p: processing.StableDiffusionProcessing, output):
results = processing_vae.vae_decode(
latents = output.images,
model = model,
full_quality = p.full_quality,
vae_type = p.vae_type,
width = width,
height = height,
frames = frames,
modules/processing_helpers.py
+11 -17
View File
@@ -201,7 +201,7 @@ def create_random_tensors(shape, seeds, subseeds=None, subseed_strength=0.0, see
return x
def decode_first_stage(model, x, full_quality=True):
def decode_first_stage(model, x):
if not shared.opts.keep_incomplete and (shared.state.skipped or shared.state.interrupted):
shared.log.debug(f'Decode VAE: skipped={shared.state.skipped} interrupted={shared.state.interrupted}')
x_sample = torch.zeros((len(x), 3, x.shape[2] * 8, x.shape[3] * 8), dtype=devices.dtype_vae, device=devices.device)
@@ -210,20 +210,14 @@ def decode_first_stage(model, x, full_quality=True):
shared.state.job = 'VAE'
with devices.autocast(disable = x.dtype==devices.dtype_vae):
try:
if full_quality:
if hasattr(model, 'decode_first_stage'):
# x_sample = model.decode_first_stage(x) * 0.5 + 0.5
x_sample = model.decode_first_stage(x)
elif hasattr(model, 'vae'):
x_sample = processing_vae.vae_decode(latents=x, model=model, output_type='np', full_quality=full_quality)
else:
x_sample = x
shared.log.error('Decode VAE unknown model')
if hasattr(model, 'decode_first_stage'):
# x_sample = model.decode_first_stage(x) * 0.5 + 0.5
x_sample = model.decode_first_stage(x)
elif hasattr(model, 'vae'):
x_sample = processing_vae.vae_decode(latents=x, model=model, output_type='np')
else:
from modules import sd_vae_taesd
x_sample = torch.zeros((len(x), 3, x.shape[2] * 8, x.shape[3] * 8), dtype=devices.dtype_vae, device=devices.device)
for i in range(len(x_sample)):
x_sample[i] = sd_vae_taesd.decode(x[i]) * 0.5 + 0.5
x_sample = x
shared.log.error('Decode VAE unknown model')
except Exception as e:
x_sample = x
shared.log.error(f'Decode VAE: {e}')
@@ -407,7 +401,7 @@ def resize_init_images(p):
def resize_hires(p, latents): # input=latents output=pil if not latent_upscaler else latent
if not torch.is_tensor(latents):
shared.log.warning('Hires: input is not tensor')
first_pass_images = processing_vae.vae_decode(latents=latents, model=shared.sd_model, full_quality=p.full_quality, output_type='pil', width=p.width, height=p.height)
first_pass_images = processing_vae.vae_decode(latents=latents, model=shared.sd_model, vae_type=p.vae_type, output_type='pil', width=p.width, height=p.height)
return first_pass_images
if (p.hr_upscale_to_x == 0 or p.hr_upscale_to_y == 0) and hasattr(p, 'init_hr'):
@@ -418,7 +412,7 @@ def resize_hires(p, latents): # input=latents output=pil if not latent_upscaler
resized_image = images.resize_image(p.hr_resize_mode, latents, p.hr_upscale_to_x, p.hr_upscale_to_y, upscaler_name=p.hr_upscaler, context=p.hr_resize_context)
return resized_image
first_pass_images = processing_vae.vae_decode(latents=latents, model=shared.sd_model, full_quality=p.full_quality, output_type='pil', width=p.width, height=p.height)
first_pass_images = processing_vae.vae_decode(latents=latents, model=shared.sd_model, vae_type=p.vae_type, output_type='pil', width=p.width, height=p.height)
resized_images = []
for img in first_pass_images:
resized_image = images.resize_image(p.hr_resize_mode, img, p.hr_upscale_to_x, p.hr_upscale_to_y, upscaler_name=p.hr_upscaler, context=p.hr_resize_context)
@@ -561,7 +555,7 @@ def save_intermediate(p, latents, suffix):
for i in range(len(latents)):
from modules.processing import create_infotext
info=create_infotext(p, p.all_prompts, p.all_seeds, p.all_subseeds, [], iteration=p.iteration, position_in_batch=i)
decoded = processing_vae.vae_decode(latents=latents, model=shared.sd_model, output_type='pil', full_quality=p.full_quality, width=p.width, height=p.height)
decoded = processing_vae.vae_decode(latents=latents, model=shared.sd_model, output_type='pil', vae_type=p.vae_type, width=p.width, height=p.height)
for j in range(len(decoded)):
images.save_image(decoded[j], path=p.outpath_samples, basename="", seed=p.seeds[i], prompt=p.prompts[i], extension=shared.opts.samples_format, info=info, p=p, suffix=suffix)
modules/processing_info.py
+4 -1
View File
@@ -58,7 +58,6 @@ def create_infotext(p: StableDiffusionProcessing, all_prompts=None, all_seeds=No
"Batch": f'{p.n_iter}x{p.batch_size}' if p.n_iter > 1 or p.batch_size > 1 else None,
"Model": None if (not shared.opts.add_model_name_to_info) or (not shared.sd_model.sd_checkpoint_info.model_name) else shared.sd_model.sd_checkpoint_info.model_name.replace(',', '').replace(':', ''),
"Model hash": getattr(p, 'sd_model_hash', None if (not shared.opts.add_model_hash_to_info) or (not shared.sd_model.sd_model_hash) else shared.sd_model.sd_model_hash),
"VAE": (None if not shared.opts.add_model_name_to_info or sd_vae.loaded_vae_file is None else os.path.splitext(os.path.basename(sd_vae.loaded_vae_file))[0]) if p.full_quality else 'TAESD',
"Refiner prompt": p.refiner_prompt if len(p.refiner_prompt) > 0 else None,
"Refiner negative": p.refiner_negative if len(p.refiner_negative) > 0 else None,
"Styles": "; ".join(p.styles) if p.styles is not None and len(p.styles) > 0 else None,
@@ -71,6 +70,10 @@ def create_infotext(p: StableDiffusionProcessing, all_prompts=None, all_seeds=No
"Comment": comment,
"Operations": '; '.join(ops).replace('"', '') if len(p.ops) > 0 else 'none',
}
if p.vae_type == 'Full':
args["VAE"] = (None if not shared.opts.add_model_name_to_info or sd_vae.loaded_vae_file is None else os.path.splitext(os.path.basename(sd_vae.loaded_vae_file))[0])
elif p.vae_type == 'Tiny':
args["VAE"] = 'TAESD'
if shared.opts.add_model_name_to_info and getattr(shared.sd_model, 'sd_checkpoint_info', None) is not None:
args["Model"] = shared.sd_model.sd_checkpoint_info.model_name.replace(',', '').replace(':', '')
if shared.opts.add_model_hash_to_info and getattr(shared.sd_model, 'sd_model_hash', None) is not None:
modules/processing_original.py
+5 -5
View File
@@ -49,7 +49,7 @@ def process_original(p: processing.StableDiffusionProcessing):
c = get_conds_with_caching(prompt_parser.get_multicond_learned_conditioning, p.prompts, p.steps * step_multiplier, cached_c)
with devices.without_autocast() if devices.unet_needs_upcast else devices.autocast():
samples_ddim = p.sample(conditioning=c, unconditional_conditioning=uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
x_samples_ddim = [processing.decode_first_stage(p.sd_model, samples_ddim[i:i+1].to(dtype=devices.dtype_vae), p.full_quality)[0].cpu() for i in range(samples_ddim.size(0))]
x_samples_ddim = [processing.decode_first_stage(p.sd_model, samples_ddim[i:i+1].to(dtype=devices.dtype_vae))[0].cpu() for i in range(samples_ddim.size(0))]
try:
for x in x_samples_ddim:
devices.test_for_nans(x, "vae")
@@ -60,7 +60,7 @@ def process_original(p: processing.StableDiffusionProcessing):
devices.dtype_vae = torch.bfloat16
vae_file, vae_source = sd_vae.resolve_vae(p.sd_model.sd_model_checkpoint)
sd_vae.load_vae(p.sd_model, vae_file, vae_source)
x_samples_ddim = [processing.decode_first_stage(p.sd_model, samples_ddim[i:i+1].to(dtype=devices.dtype_vae), p.full_quality)[0].cpu() for i in range(samples_ddim.size(0))]
x_samples_ddim = [processing.decode_first_stage(p.sd_model, samples_ddim[i:i+1].to(dtype=devices.dtype_vae))[0].cpu() for i in range(samples_ddim.size(0))]
for x in x_samples_ddim:
devices.test_for_nans(x, "vae")
else:
@@ -90,7 +90,7 @@ def sample_txt2img(p: processing.StableDiffusionProcessingTxt2Img, conditioning,
target_height = p.hr_upscale_to_y
decoded_samples = None
if shared.opts.samples_save and shared.opts.save_images_before_highres_fix and not p.do_not_save_samples:
decoded_samples = decode_first_stage(p.sd_model, samples.to(dtype=devices.dtype_vae), p.full_quality)
decoded_samples = decode_first_stage(p.sd_model, samples.to(dtype=devices.dtype_vae))
decoded_samples = torch.clamp((decoded_samples + 1.0) / 2.0, min=0.0, max=1.0)
for i, x_sample in enumerate(decoded_samples):
x_sample = validate_sample(x_sample)
@@ -107,13 +107,13 @@ def sample_txt2img(p: processing.StableDiffusionProcessingTxt2Img, conditioning,
shared.state.job = 'Upscale'
samples = images.resize_image(1, samples, target_width, target_height, upscaler_name=p.hr_upscaler)
if getattr(p, "inpainting_mask_weight", shared.opts.inpainting_mask_weight) < 1.0:
image_conditioning = img2img_image_conditioning(p, decode_first_stage(p.sd_model, samples.to(dtype=devices.dtype_vae), p.full_quality), samples)
image_conditioning = img2img_image_conditioning(p, decode_first_stage(p.sd_model, samples.to(dtype=devices.dtype_vae)), samples)
else:
image_conditioning = txt2img_image_conditioning(p, samples.to(dtype=devices.dtype_vae))
else:
shared.state.job = 'Upscale'
if decoded_samples is None:
decoded_samples = decode_first_stage(p.sd_model, samples.to(dtype=devices.dtype_vae), p.full_quality)
decoded_samples = decode_first_stage(p.sd_model, samples.to(dtype=devices.dtype_vae))
decoded_samples = torch.clamp((decoded_samples + 1.0) / 2.0, min=0.0, max=1.0)
batch_images = []
for _i, x_sample in enumerate(decoded_samples):
modules/processing_vae.py
+16 -7
View File
@@ -17,9 +17,9 @@ def create_latents(image, p, dtype=None, device=None):
if image is None:
return image
elif isinstance(image, Image.Image):
latents = vae_encode(image, model=shared.sd_model, full_quality=p.full_quality)
latents = vae_encode(image, model=shared.sd_model, vae_type=p.vae_type)
elif isinstance(image, list):
latents = [vae_encode(i, model=shared.sd_model, full_quality=p.full_quality).squeeze(dim=0) for i in image]
latents = [vae_encode(i, model=shared.sd_model, vae_type=p.vae_type).squeeze(dim=0) for i in image]
latents = torch.stack(latents, dim=0).to(shared.device)
else:
shared.log.warning(f'Latents: input type: {type(image)} {image}')
@@ -230,7 +230,7 @@ def taesd_vae_encode(image):
return encoded
def vae_decode(latents, model, output_type='np', full_quality=True, width=None, height=None, frames=None):
def vae_decode(latents, model, output_type='np', vae_type='Full', width=None, height=None, frames=None):
t0 = time.time()
model = model or shared.sd_model
if not hasattr(model, 'vae') and hasattr(model, 'pipe'):
@@ -238,6 +238,15 @@ def vae_decode(latents, model, output_type='np', full_quality=True, width=None,
if latents is None or not torch.is_tensor(latents): # already decoded
return latents
prev_job = shared.state.job
if vae_type == 'Remote':
shared.state.job = 'Remote VAE'
from modules.sd_vae_remote import remote_decode
images = remote_decode(latents=latents, width=width, height=height)
shared.state.job = prev_job
if images is not None and len(images) > 0:
return images
shared.state.job = 'VAE'
if latents.shape[0] == 0:
shared.log.error(f'VAE nothing to decode: {latents.shape}')
@@ -261,7 +270,7 @@ def vae_decode(latents, model, output_type='np', full_quality=True, width=None,
if latents.shape[-1] <= 4: # not a latent, likely an image
decoded = latents.float().cpu().numpy()
elif full_quality and hasattr(model, "vae"):
elif vae_type == 'Full' and hasattr(model, "vae"):
decoded = full_vae_decode(latents=latents, model=model)
elif hasattr(model, "vqgan"):
decoded = full_vqgan_decode(latents=latents, model=model)
@@ -296,7 +305,7 @@ def vae_decode(latents, model, output_type='np', full_quality=True, width=None,
return imgs
def vae_encode(image, model, full_quality=True): # pylint: disable=unused-variable
def vae_encode(image, model, vae_type='Full'): # pylint: disable=unused-variable
if shared.state.interrupted or shared.state.skipped:
return []
if not hasattr(model, 'vae') and hasattr(model, 'pipe'):
@@ -305,7 +314,7 @@ def vae_encode(image, model, full_quality=True): # pylint: disable=unused-variab
shared.log.error('VAE not found in model')
return []
tensor = TF.to_tensor(image.convert("RGB")).unsqueeze(0).to(devices.device, devices.dtype_vae)
if full_quality:
if vae_type == 'Full':
tensor = tensor * 2 - 1
latents = full_vae_encode(image=tensor, model=shared.sd_model)
else:
@@ -321,7 +330,7 @@ def reprocess(gallery):
if latent is None or gallery is None:
return None
shared.log.info(f'Reprocessing: latent={latent.shape}')
reprocessed = vae_decode(latent, shared.sd_model, output_type='pil', full_quality=True)
reprocessed = vae_decode(latent, shared.sd_model, output_type='pil')
outputs = []
for i0, i1 in zip(gallery, reprocessed):
if isinstance(i1, np.ndarray):
modules/sd_vae_remote.py
+50
View File
@@ -0,0 +1,50 @@
import io
import time
import base64
import torch
import requests
from PIL import Image
from safetensors.torch import _tobytes
hf_endpoints = {
'sd': 'https://lqmfdhmzmy4dw51z.us-east-1.aws.endpoints.huggingface.cloud',
'sdxl': 'https://m5fxqwyk0r3uu79o.us-east-1.aws.endpoints.huggingface.cloud',
'f1': 'https://zy1z7fzxpgtltg06.us-east-1.aws.endpoints.huggingface.cloud',
}
def remote_decode(latents: torch.Tensor, width: int = 0, height: int = 0, model_type: str = None) -> Image.Image:
from modules import devices, shared, errors
images = []
model_type = model_type or shared.sd_model_type
url = hf_endpoints.get(model_type, None)
if url is None:
shared.log.error(f'Decode: type="remote" type={model_type} unsuppported')
return images
t0 = time.time()
latents = latents.unsqueeze(0) if len(latents.shape) == 3 else latents
for i in range(latents.shape[0]):
try:
latent = latents[i].detach().clone().to(device=devices.cpu, dtype=devices.dtype).unsqueeze(0)
encoded = base64.b64encode(_tobytes(latent, "inputs")).decode("utf-8")
params = {"shape": list(latent.shape), "dtype": str(latent.dtype).split(".", maxsplit=1)[-1]}
if (model_type == 'f1') and (width > 0) and (height > 0):
params['width'] = width
params['height'] = height
response = requests.post(
url=url,
json={"inputs": encoded, "parameters": params},
headers={"Content-Type": "application/json", "Accept": "image/jpeg"},
timeout=60,
)
if not response.ok:
shared.log.error(f'Decode: type="remote" model={model_type} code={response.status_code} {response.json()}')
else:
images.append(Image.open(io.BytesIO(response.content)))
except Exception as e:
shared.log.error(f'Decode: type="remote" model={model_type} {e}')
errors.display(e, 'VAE')
t1 = time.time()
shared.log.debug(f'Decode: type="remote" model={model_type} args={params} images={images} time={t1-t0:.3f}s')
return images
modules/txt2img.py
+2 -2
View File
@@ -11,7 +11,7 @@ debug('Trace: PROCESS')
def txt2img(id_task, state,
prompt, negative_prompt, prompt_styles,
steps, sampler_index, hr_sampler_index,
full_quality, tiling, hidiffusion,
vae_type, tiling, hidiffusion,
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength,
n_iter, batch_size,
cfg_scale, image_cfg_scale, diffusers_guidance_rescale, pag_scale, pag_adaptive, cfg_end,
@@ -64,7 +64,7 @@ def txt2img(id_task, state,
clip_skip=clip_skip,
width=width,
height=height,
full_quality=full_quality,
vae_type=vae_type,
detailer_enabled=detailer_enabled,
detailer_prompt=detailer_prompt,
detailer_negative=detailer_negative,
modules/ui_control.py
+3 -3
View File
@@ -161,7 +161,7 @@ def create_ui(_blocks: gr.Blocks=None):
mask_controls = masking.create_segment_ui()
full_quality, tiling, hidiffusion, cfg_scale, clip_skip, image_cfg_scale, guidance_rescale, pag_scale, pag_adaptive, cfg_end = ui_sections.create_advanced_inputs('control')
vae_type, tiling, hidiffusion, cfg_scale, clip_skip, image_cfg_scale, guidance_rescale, pag_scale, pag_adaptive, cfg_end = ui_sections.create_advanced_inputs('control')
hdr_mode, hdr_brightness, hdr_color, hdr_sharpen, hdr_clamp, hdr_boundary, hdr_threshold, hdr_maximize, hdr_max_center, hdr_max_boundry, hdr_color_picker, hdr_tint_ratio = ui_sections.create_correction_inputs('control')
with gr.Accordion(open=False, label="Video", elem_id="control_video", elem_classes=["small-accordion"]):
@@ -561,7 +561,7 @@ def create_ui(_blocks: gr.Blocks=None):
prompt, negative, styles,
steps, sampler_index,
seed, subseed, subseed_strength, seed_resize_from_h, seed_resize_from_w,
cfg_scale, clip_skip, image_cfg_scale, guidance_rescale, pag_scale, pag_adaptive, cfg_end, full_quality, tiling, hidiffusion,
cfg_scale, clip_skip, image_cfg_scale, guidance_rescale, pag_scale, pag_adaptive, cfg_end, vae_type, tiling, hidiffusion,
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength,
hdr_mode, hdr_brightness, hdr_color, hdr_sharpen, hdr_clamp, hdr_boundary, hdr_threshold, hdr_maximize, hdr_max_center, hdr_max_boundry, hdr_color_picker, hdr_tint_ratio,
resize_mode_before, resize_name_before, resize_context_before, width_before, height_before, scale_by_before, selected_scale_tab_before,
@@ -646,7 +646,7 @@ def create_ui(_blocks: gr.Blocks=None):
(image_cfg_scale, "Image CFG scale"),
(image_cfg_scale, "Hires CFG scale"),
(guidance_rescale, "CFG rescale"),
(full_quality, "Full quality"),
(vae_type, "VAE type"),
(tiling, "Tiling"),
(hidiffusion, "HiDiffusion"),
# detailer
modules/ui_img2img.py
+3 -3
View File
@@ -131,7 +131,7 @@ def create_ui():
denoising_strength = gr.Slider(minimum=0.0, maximum=0.99, step=0.01, label='Denoising strength', value=0.30, elem_id="img2img_denoising_strength")
refiner_start = gr.Slider(minimum=0.0, maximum=1.0, step=0.05, label='Denoise start', value=0.0, elem_id="img2img_refiner_start")
full_quality, tiling, hidiffusion, cfg_scale, clip_skip, image_cfg_scale, diffusers_guidance_rescale, pag_scale, pag_adaptive, cfg_end = ui_sections.create_advanced_inputs('img2img')
vae_type, tiling, hidiffusion, cfg_scale, clip_skip, image_cfg_scale, diffusers_guidance_rescale, pag_scale, pag_adaptive, cfg_end = ui_sections.create_advanced_inputs('img2img')
hdr_mode, hdr_brightness, hdr_color, hdr_sharpen, hdr_clamp, hdr_boundary, hdr_threshold, hdr_maximize, hdr_max_center, hdr_max_boundry, hdr_color_picker, hdr_tint_ratio = ui_sections.create_correction_inputs('img2img')
enable_hr, hr_sampler_index, hr_denoising_strength, hr_resize_mode, hr_resize_context, hr_upscaler, hr_force, hr_second_pass_steps, hr_scale, hr_resize_x, hr_resize_y, refiner_steps, hr_refiner_start, refiner_prompt, refiner_negative = ui_sections.create_hires_inputs('txt2img')
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength = shared.yolo.ui('img2img')
@@ -175,7 +175,7 @@ def create_ui():
sampler_index,
mask_blur, mask_alpha,
inpainting_fill,
full_quality, tiling, hidiffusion,
vae_type, tiling, hidiffusion,
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength,
batch_count, batch_size,
cfg_scale, image_cfg_scale,
@@ -261,7 +261,7 @@ def create_ui():
(image_cfg_scale, "Hires CFG scale"),
(clip_skip, "Clip skip"),
(diffusers_guidance_rescale, "CFG rescale"),
(full_quality, "Full quality"),
(vae_type, "VAE type"),
(tiling, "Tiling"),
(hidiffusion, "HiDiffusion"),
# detailer
modules/ui_sections.py
+4 -3
View File
@@ -170,8 +170,9 @@ def create_advanced_inputs(tab, base=True):
with gr.Accordion(open=False, label="Advanced", elem_id=f"{tab}_advanced", elem_classes=["small-accordion"]):
with gr.Group():
with gr.Row(elem_id=f"{tab}_advanced_options"):
full_quality = gr.Checkbox(label='Full quality', value=True, elem_id=f"{tab}_full_quality")
tiling = gr.Checkbox(label='Tiling', value=False, elem_id=f"{tab}_tiling")
vae_type = gr.Dropdown(label='VAE type', choices=['Full', 'Tiny', 'Remote'], value='Full', elem_id=f"{tab}_vae_type")
with gr.Row(elem_id=f"{tab}_advanced_options"):
tiling = gr.Checkbox(label='Texture tiling', value=False, elem_id=f"{tab}_tiling")
hidiffusion = gr.Checkbox(label='HiDiffusion', value=False, elem_id=f"{tab}_hidiffusion")
if base:
cfg_scale, cfg_end = create_cfg_inputs(tab)
@@ -185,7 +186,7 @@ def create_advanced_inputs(tab, base=True):
diffusers_pag_adaptive = gr.Slider(minimum=0.0, maximum=1.0, step=0.05, label='Adaptive scaling', value=0.5, elem_id=f"{tab}_pag_adaptive", visible=shared.native)
with gr.Row():
clip_skip = gr.Slider(label='CLIP skip', value=1, minimum=0, maximum=12, step=0.1, elem_id=f"{tab}_clip_skip", interactive=True)
return full_quality, tiling, hidiffusion, cfg_scale, clip_skip, image_cfg_scale, diffusers_guidance_rescale, diffusers_pag_scale, diffusers_pag_adaptive, cfg_end
return vae_type, tiling, hidiffusion, cfg_scale, clip_skip, image_cfg_scale, diffusers_guidance_rescale, diffusers_pag_scale, diffusers_pag_adaptive, cfg_end
def create_correction_inputs(tab):
modules/ui_txt2img.py
+3 -3
View File
@@ -44,7 +44,7 @@ def create_ui():
with gr.Accordion(open=False, label="Samplers", elem_classes=["small-accordion"], elem_id="txt2img_sampler_group"):
ui_sections.create_sampler_options('txt2img')
seed, reuse_seed, subseed, reuse_subseed, subseed_strength, seed_resize_from_h, seed_resize_from_w = ui_sections.create_seed_inputs('txt2img')
full_quality, tiling, hidiffusion, _cfg_scale, clip_skip, image_cfg_scale, diffusers_guidance_rescale, pag_scale, pag_adaptive, _cfg_end = ui_sections.create_advanced_inputs('txt2img', base=False)
vae_type, tiling, hidiffusion, _cfg_scale, clip_skip, image_cfg_scale, diffusers_guidance_rescale, pag_scale, pag_adaptive, _cfg_end = ui_sections.create_advanced_inputs('txt2img', base=False)
hdr_mode, hdr_brightness, hdr_color, hdr_sharpen, hdr_clamp, hdr_boundary, hdr_threshold, hdr_maximize, hdr_max_center, hdr_max_boundry, hdr_color_picker, hdr_tint_ratio = ui_sections.create_correction_inputs('txt2img')
enable_hr, hr_sampler_index, denoising_strength, hr_resize_mode, hr_resize_context, hr_upscaler, hr_force, hr_second_pass_steps, hr_scale, hr_resize_x, hr_resize_y, refiner_steps, refiner_start, refiner_prompt, refiner_negative = ui_sections.create_hires_inputs('txt2img')
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength = shared.yolo.ui('txt2img')
@@ -64,7 +64,7 @@ def create_ui():
dummy_component, state,
txt2img_prompt, txt2img_negative_prompt, txt2img_prompt_styles,
steps, sampler_index, hr_sampler_index,
full_quality, tiling, hidiffusion,
vae_type, tiling, hidiffusion,
detailer_enabled, detailer_prompt, detailer_negative, detailer_steps, detailer_strength,
batch_count, batch_size,
cfg_scale, image_cfg_scale, diffusers_guidance_rescale, pag_scale, pag_adaptive, cfg_end,
@@ -122,7 +122,7 @@ def create_ui():
(image_cfg_scale, "Image CFG scale"),
(image_cfg_scale, "Hires CFG scale"),
(diffusers_guidance_rescale, "CFG rescale"),
(full_quality, "Full quality"),
(vae_type, "VAE type"),
(tiling, "Tiling"),
(hidiffusion, "HiDiffusion"),
# detailer