Mirrored_Repos/InvokeAI
1
0
Fork 0
mirror of https://github.com/invoke-ai/InvokeAI synced 2024-08-30 20:32:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix Inpainting Issues (#3744)

Browse Source 
- fix: Inpaint not working with some schedulers: Resolves #3732
- fix: LoRA's not working at all while inpainting.
This commit is contained in:
blessedcoolant 2023-07-13 23:42:44 +12:00 committed by GitHub
commit d4ec8873f7
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 47 additions and 45 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

90
invokeai/app/invocations/generate.py
View File

@ -154,40 +154,42 @@ class InpaintInvocation(BaseInvocation):
@contextmanager @contextmanager
def load_model_old_way(self, context, scheduler): def load_model_old_way(self, context, scheduler):
def _lora_loader():
for lora in self.unet.loras:
lora_info = context.services.model_manager.get_model(
**lora.dict(exclude={"weight"}))
yield (lora_info.context.model, lora.weight)
del lora_info
return
unet_info = context.services.model_manager.get_model(**self.unet.unet.dict()) unet_info = context.services.model_manager.get_model(**self.unet.unet.dict())
vae_info = context.services.model_manager.get_model(**self.vae.vae.dict()) vae_info = context.services.model_manager.get_model(**self.vae.vae.dict())
#unet = unet_info.context.model with vae_info as vae,\
#vae = vae_info.context.model ModelPatcher.apply_lora_unet(unet_info.context.model, _lora_loader()),\
unet_info as unet:
with ExitStack() as stack: device = context.services.model_manager.mgr.cache.execution_device
loras = [(stack.enter_context(context.services.model_manager.get_model(**lora.dict(exclude={"weight"}))), lora.weight) for lora in self.unet.loras] dtype = context.services.model_manager.mgr.cache.precision
with vae_info as vae,\ pipeline = StableDiffusionGeneratorPipeline(
unet_info as unet,\ vae=vae,
ModelPatcher.apply_lora_unet(unet, loras): text_encoder=None,
tokenizer=None,
unet=unet,
scheduler=scheduler,
safety_checker=None,
feature_extractor=None,
requires_safety_checker=False,
precision="float16" if dtype == torch.float16 else "float32",
execution_device=device,
)
device = context.services.model_manager.mgr.cache.execution_device yield OldModelInfo(
dtype = context.services.model_manager.mgr.cache.precision name=self.unet.unet.model_name,
hash="<NO-HASH>",
pipeline = StableDiffusionGeneratorPipeline( model=pipeline,
vae=vae, )
text_encoder=None,
tokenizer=None,
unet=unet,
scheduler=scheduler,
safety_checker=None,
feature_extractor=None,
requires_safety_checker=False,
precision="float16" if dtype == torch.float16 else "float32",
execution_device=device,
)
yield OldModelInfo(
name=self.unet.unet.model_name,
hash="<NO-HASH>",
model=pipeline,
)
def invoke(self, context: InvocationContext) -> ImageOutput: def invoke(self, context: InvocationContext) -> ImageOutput:
image = ( image = (
@ -226,21 +228,21 @@ class InpaintInvocation(BaseInvocation):
), # Shorthand for passing all of the parameters above manually ), # Shorthand for passing all of the parameters above manually
) )
# Outputs is an infinite iterator that will return a new InvokeAIGeneratorOutput object # Outputs is an infinite iterator that will return a new InvokeAIGeneratorOutput object
# each time it is called. We only need the first one. # each time it is called. We only need the first one.
generator_output = next(outputs) generator_output = next(outputs)
image_dto = context.services.images.create( image_dto = context.services.images.create(
image=generator_output.image, image=generator_output.image,
image_origin=ResourceOrigin.INTERNAL, image_origin=ResourceOrigin.INTERNAL,
image_category=ImageCategory.GENERAL, image_category=ImageCategory.GENERAL,
session_id=context.graph_execution_state_id, session_id=context.graph_execution_state_id,
node_id=self.id, node_id=self.id,
is_intermediate=self.is_intermediate, is_intermediate=self.is_intermediate,
) )
return ImageOutput( return ImageOutput(
image=ImageField(image_name=image_dto.image_name), image=ImageField(image_name=image_dto.image_name),
width=image_dto.width, width=image_dto.width,
height=image_dto.height, height=image_dto.height,
) )

2
invokeai/backend/stable_diffusion/diffusers_pipeline.py
View File

@ -127,7 +127,7 @@ class AddsMaskGuidance:
def _t_for_field(self, field_name: str, t): def _t_for_field(self, field_name: str, t):
if field_name == "pred_original_sample": if field_name == "pred_original_sample":
return torch.zeros_like(t, dtype=t.dtype) # it represents t=0 return self.scheduler.timesteps[-1]
return t return t
def apply_mask(self, latents: torch.Tensor, t) -> torch.Tensor: def apply_mask(self, latents: torch.Tensor, t) -> torch.Tensor: