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

Consolidate latents_from_embeddings(...) and generate_latents_from_embeddings(...) into a single function.

Browse Source
This commit is contained in:
Ryan Dick 2024-06-13 12:24:43 -04:00 committed by Kent Keirsey
parent 80a67572f1
commit f604575862
1 changed files with 64 additions and 86 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
invokeai/backend/stable_diffusion/diffusers_pipeline.py
View File

@ -325,44 +325,6 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
mask_guidance = AddsMaskGuidance(mask, orig_latents, self.scheduler, noise, is_gradient_mask) mask_guidance = AddsMaskGuidance(mask, orig_latents, self.scheduler, noise, is_gradient_mask)
try: try:
latents = self.generate_latents_from_embeddings(
latents,
timesteps,
conditioning_data,
scheduler_step_kwargs=scheduler_step_kwargs,
mask_guidance=mask_guidance,
control_data=control_data,
ip_adapter_data=ip_adapter_data,
t2i_adapter_data=t2i_adapter_data,
callback=callback,
)
finally:
self.invokeai_diffuser.model_forward_callback = self._unet_forward
# restore unmasked part after the last step is completed
# in-process masking happens before each step
if mask is not None:
if is_gradient_mask:
latents = torch.where(mask > 0, latents, orig_latents)
else:
latents = torch.lerp(
orig_latents, latents.to(dtype=orig_latents.dtype), mask.to(dtype=orig_latents.dtype)
)
return latents
def generate_latents_from_embeddings(
self,
latents: torch.Tensor,
timesteps: torch.Tensor,
conditioning_data: TextConditioningData,
scheduler_step_kwargs: dict[str, Any],
callback: Callable[[PipelineIntermediateState], None],
mask_guidance: AddsMaskGuidance | None = None,
control_data: list[ControlNetData] | None = None,
ip_adapter_data: Optional[list[IPAdapterData]] = None,
t2i_adapter_data: Optional[list[T2IAdapterData]] = None,
) -> torch.Tensor:
self._adjust_memory_efficient_attention(latents) self._adjust_memory_efficient_attention(latents)
batch_size = latents.shape[0] batch_size = latents.shape[0]
@ -379,7 +341,10 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
if use_ip_adapter or use_regional_prompting: if use_ip_adapter or use_regional_prompting:
ip_adapters: Optional[List[UNetIPAdapterData]] = ( ip_adapters: Optional[List[UNetIPAdapterData]] = (
[{"ip_adapter": ipa.ip_adapter_model, "target_blocks": ipa.target_blocks} for ipa in ip_adapter_data] [
{"ip_adapter": ipa.ip_adapter_model, "target_blocks": ipa.target_blocks}
for ipa in ip_adapter_data
]
if use_ip_adapter if use_ip_adapter
else None else None
) )
@ -425,6 +390,19 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
) )
) )
finally:
self.invokeai_diffuser.model_forward_callback = self._unet_forward
# restore unmasked part after the last step is completed
# in-process masking happens before each step
if mask is not None:
if is_gradient_mask:
latents = torch.where(mask > 0, latents, orig_latents)
else:
latents = torch.lerp(
orig_latents, latents.to(dtype=orig_latents.dtype), mask.to(dtype=orig_latents.dtype)
)
return latents return latents
@torch.inference_mode() @torch.inference_mode()