OPENGIT/stable-diffusion-webui
1
0
Fork 0
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git synced 2025-02-10 23:52:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

Added image conditioning to latent upscale.

Browse Source 
Only comuted  if the mask weight is not 1.0 to avoid extra memory.
Also includes some code cleanup.
This commit is contained in:
random_thoughtss 2022-10-29 10:35:51 -07:00
parent 44ab954fab
commit 6e2ce4e735
1 changed files with 11 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
modules/processing.py
View File

@ -134,11 +134,7 @@ class StableDiffusionProcessing():
# Dummy zero conditioning if we're not using inpainting model. # Dummy zero conditioning if we're not using inpainting model.
# Still takes up a bit of memory, but no encoder call. # Still takes up a bit of memory, but no encoder call.
# Pretty sure we can just make this a 1x1 image since its not going to be used besides its batch size. # Pretty sure we can just make this a 1x1 image since its not going to be used besides its batch size.
return torch.zeros( return x.new_zeros(x.shape[0], 5, 1, 1)
x.shape[0], 5, 1, 1,
dtype=x.dtype,
device=x.device
)
height = height or self.height height = height or self.height
width = width or self.width width = width or self.width
@ -156,11 +152,7 @@ class StableDiffusionProcessing():
def img2img_image_conditioning(self, source_image, latent_image, image_mask = None): def img2img_image_conditioning(self, source_image, latent_image, image_mask = None):
if self.sampler.conditioning_key not in {'hybrid', 'concat'}: if self.sampler.conditioning_key not in {'hybrid', 'concat'}:
# Dummy zero conditioning if we're not using inpainting model. # Dummy zero conditioning if we're not using inpainting model.
return torch.zeros( return latent_image.new_zeros(latent_image.shape[0], 5, 1, 1)
latent_image.shape[0], 5, 1, 1,
dtype=latent_image.dtype,
device=latent_image.device
)
# Handle the different mask inputs # Handle the different mask inputs
if image_mask is not None: if image_mask is not None:
@ -174,11 +166,10 @@ class StableDiffusionProcessing():
# Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0 # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0
conditioning_mask = torch.round(conditioning_mask) conditioning_mask = torch.round(conditioning_mask)
else: else:
conditioning_mask = torch.ones(1, 1, *source_image.shape[-2:]) conditioning_mask = source_image.new_ones(1, 1, *source_image.shape[-2:])
# Create another latent image, this time with a masked version of the original input. # Create another latent image, this time with a masked version of the original input.
# Smoothly interpolate between the masked and unmasked latent conditioning image using a parameter. # Smoothly interpolate between the masked and unmasked latent conditioning image using a parameter.
conditioning_mask = conditioning_mask.to(source_image.device)
conditioning_image = torch.lerp( conditioning_image = torch.lerp(
source_image, source_image,
source_image * (1.0 - conditioning_mask), source_image * (1.0 - conditioning_mask),
@ -653,7 +644,13 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
if opts.use_scale_latent_for_hires_fix: if opts.use_scale_latent_for_hires_fix:
samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), mode="bilinear") samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), mode="bilinear")
image_conditioning = self.txt2img_image_conditioning(samples)
# Avoid making the inpainting conditioning unless necessary as
# this does need some extra compute to decode / encode the image again.
if getattr(self, "inpainting_mask_weight", shared.opts.inpainting_mask_weight) < 1.0:
image_conditioning = self.img2img_image_conditioning(decode_first_stage(self.sd_model, samples), samples)
else:
image_conditioning = self.txt2img_image_conditioning(samples)
else: else:
decoded_samples = decode_first_stage(self.sd_model, samples) decoded_samples = decode_first_stage(self.sd_model, samples)
@ -675,11 +672,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
samples = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(decoded_samples)) samples = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(decoded_samples))
image_conditioning = self.img2img_image_conditioning( image_conditioning = self.img2img_image_conditioning(decoded_samples, samples)
decoded_samples,
samples,
decoded_samples.new_ones(decoded_samples.shape[0], 1, decoded_samples.shape[2], decoded_samples.shape[3])
)
shared.state.nextjob() shared.state.nextjob()