Merge branch 'main' into enhance/invokeai-logs

- resolve conflicts
- remove unused code identified by pyflakes

invokeai/backend/__init__.py

@@ -10,7 +10,7 @@ from .generator import (
     Img2Img,
     Inpaint
 )
-from .model_management import ModelManager
+from .model_management import ModelManager, SDModelComponent
 from .safety_checker import SafetyChecker
 from .args import Args
 from .globals import Globals

invokeai/backend/model_management/__init__.py

@@ -5,6 +5,7 @@ from .convert_ckpt_to_diffusers import (
     convert_ckpt_to_diffusers,
     load_pipeline_from_original_stable_diffusion_ckpt,
 )
-from .model_manager import ModelManager
+from .model_manager import ModelManager,SDModelComponent
+
 
 

invokeai/backend/stable_diffusion/concepts_lib.py

@@ -58,7 +58,7 @@ class HuggingFaceConceptsLibrary(object):
                 self.concept_list.extend(list(local_concepts_to_add))
                 return self.concept_list
             return self.concept_list
-        else:
+        elif Globals.internet_available is True:
             try:
                 models = self.hf_api.list_models(
                     filter=ModelFilter(model_name="sd-concepts-library/")
@@ -74,6 +74,8 @@ class HuggingFaceConceptsLibrary(object):
                     "You may load .bin and .pt file(s) manually using the --embedding_directory argument."
                 )
             return self.concept_list
+        else: 
+            return self.concept_list
 
     def get_concept_model_path(self, concept_name: str) -> str:
         """

invokeai/backend/stable_diffusion/diffusers_pipeline.py

@@ -445,8 +445,15 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
     @property
     def _submodels(self) -> Sequence[torch.nn.Module]:
         module_names, _, _ = self.extract_init_dict(dict(self.config))
-        values = [getattr(self, name) for name in module_names.keys()]
-        return [m for m in values if isinstance(m, torch.nn.Module)]
+        submodels = []
+        for name in module_names.keys():
+            if hasattr(self, name):
+                value = getattr(self, name)
+            else:
+                value = getattr(self.config, name)
+            if isinstance(value, torch.nn.Module):
+                submodels.append(value)
+        return submodels
 
     def image_from_embeddings(
         self,
@@ -544,7 +551,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
             yield PipelineIntermediateState(
                 run_id=run_id,
                 step=-1,
-                timestep=self.scheduler.num_train_timesteps,
+                timestep=self.scheduler.config.num_train_timesteps,
                 latents=latents,
             )
 
@@ -915,7 +922,7 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
     @property
     def channels(self) -> int:
         """Compatible with DiffusionWrapper"""
-        return self.unet.in_channels
+        return self.unet.config.in_channels
 
     def decode_latents(self, latents):
         # Explicit call to get the vae loaded, since `decode` isn't the forward method.

invokeai/backend/stable_diffusion/diffusion/cross_attention_control.py

@@ -10,8 +10,7 @@ import diffusers
 import psutil
 import torch
 from compel.cross_attention_control import Arguments
-from diffusers.models.cross_attention import AttnProcessor
-from diffusers.models.unet_2d_condition import UNet2DConditionModel
+from diffusers.models.attention_processor import AttentionProcessor
 from torch import nn
 
 import invokeai.backend.util.logging as log
@@ -188,7 +187,7 @@ class Context:
 
 class InvokeAICrossAttentionMixin:
     """
-    Enable InvokeAI-flavoured CrossAttention calculation, which does aggressive low-memory slicing and calls
+    Enable InvokeAI-flavoured Attention calculation, which does aggressive low-memory slicing and calls
     through both to an attention_slice_wrangler and a slicing_strategy_getter for custom attention map wrangling
     and dymamic slicing strategy selection.
     """
@@ -209,7 +208,7 @@ class InvokeAICrossAttentionMixin:
         Set custom attention calculator to be called when attention is calculated
         :param wrangler: Callback, with args (module, suggested_attention_slice, dim, offset, slice_size),
         which returns either the suggested_attention_slice or an adjusted equivalent.
-            `module` is the current CrossAttention module for which the callback is being invoked.
+            `module` is the current Attention module for which the callback is being invoked.
             `suggested_attention_slice` is the default-calculated attention slice
             `dim` is -1 if the attenion map has not been sliced, or 0 or 1 for dimension-0 or dimension-1 slicing.
                 If `dim` is >= 0, `offset` and `slice_size` specify the slice start and length.
@@ -345,11 +344,11 @@ class InvokeAICrossAttentionMixin:
 def restore_default_cross_attention(
     model,
     is_running_diffusers: bool,
-    restore_attention_processor: Optional[AttnProcessor] = None,
+    restore_attention_processor: Optional[AttentionProcessor] = None,
 ):
     if is_running_diffusers:
         unet = model
-        unet.set_attn_processor(restore_attention_processor or CrossAttnProcessor())
+        unet.set_attn_processor(restore_attention_processor or AttnProcessor())
     else:
         remove_attention_function(model)
 
@@ -408,12 +407,9 @@ def override_cross_attention(model, context: Context, is_running_diffusers=False
 def get_cross_attention_modules(
     model, which: CrossAttentionType
 ) -> list[tuple[str, InvokeAICrossAttentionMixin]]:
-    from ldm.modules.attention import CrossAttention  # avoid circular import
 
     cross_attention_class: type = (
         InvokeAIDiffusersCrossAttention
-        if isinstance(model, UNet2DConditionModel)
-        else CrossAttention
     )
     which_attn = "attn1" if which is CrossAttentionType.SELF else "attn2"
     attention_module_tuples = [
@@ -550,7 +546,7 @@ def get_mem_free_total(device):
 
 
 class InvokeAIDiffusersCrossAttention(
-    diffusers.models.attention.CrossAttention, InvokeAICrossAttentionMixin
+    diffusers.models.attention.Attention, InvokeAICrossAttentionMixin
 ):
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
@@ -572,8 +568,8 @@ class InvokeAIDiffusersCrossAttention(
 """
 # base implementation
 
-class CrossAttnProcessor:
-    def __call__(self, attn: CrossAttention, hidden_states, encoder_hidden_states=None, attention_mask=None):
+class AttnProcessor:
+    def __call__(self, attn: Attention, hidden_states, encoder_hidden_states=None, attention_mask=None):
         batch_size, sequence_length, _ = hidden_states.shape
         attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length)
 
@@ -601,9 +597,9 @@ class CrossAttnProcessor:
 from dataclasses import dataclass, field
 
 import torch
-from diffusers.models.cross_attention import (
-    CrossAttention,
-    CrossAttnProcessor,
+from diffusers.models.attention_processor import (
+    Attention,
+    AttnProcessor,
     SlicedAttnProcessor,
 )
 
@@ -653,7 +649,7 @@ class SlicedSwapCrossAttnProcesser(SlicedAttnProcessor):
 
     def __call__(
         self,
-        attn: CrossAttention,
+        attn: Attention,
         hidden_states,
         encoder_hidden_states=None,
         attention_mask=None,

invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py

@@ -5,7 +5,7 @@ from typing import Any, Callable, Dict, Optional, Union
 
 import numpy as np
 import torch
-from diffusers.models.cross_attention import AttnProcessor
+from diffusers.models.attention_processor import AttentionProcessor
 from typing_extensions import TypeAlias
 
 import invokeai.backend.util.logging as log
@@ -102,7 +102,7 @@ class InvokeAIDiffuserComponent:
 
     def override_cross_attention(
         self, conditioning: ExtraConditioningInfo, step_count: int
-    ) -> Dict[str, AttnProcessor]:
+    ) -> Dict[str, AttentionProcessor]:
         """
         setup cross attention .swap control. for diffusers this replaces the attention processor, so
         the previous attention processor is returned so that the caller can restore it later.
@@ -119,7 +119,7 @@ class InvokeAIDiffuserComponent:
         )
 
     def restore_default_cross_attention(
-        self, restore_attention_processor: Optional["AttnProcessor"] = None
+        self, restore_attention_processor: Optional["AttentionProcessor"] = None
     ):
         self.conditioning = None
         self.cross_attention_control_context = None