Fix loading diffusers ti

invokeai/app/invocations/compel.py

@@ -9,6 +9,7 @@ from compel.prompt_parser import (Blend, Conjunction,
                                   FlattenedPrompt, Fragment)
 from pydantic import BaseModel, Field
 
+from ...backend.model_management.models import ModelNotFoundException
 from ...backend.model_management import BaseModelType, ModelType, SubModelType
 from ...backend.model_management.lora import ModelPatcher
 from ...backend.stable_diffusion.diffusion import InvokeAIDiffuserComponent
@@ -86,7 +87,7 @@ class CompelInvocation(BaseInvocation):
                         model_type=ModelType.TextualInversion,
                     ).context.model
                 )
-            except Exception:
+            except ModelNotFoundException:
                 # print(e)
                 #import traceback
                 #print(traceback.format_exc())

invokeai/backend/model_management/lora.py

@@ -655,6 +655,9 @@ class TextualInversionModel:
         else:
             result.embedding = next(iter(state_dict.values()))
 
+            if len(result.embedding.shape) == 1:
+                result.embedding = result.embedding.unsqueeze(0)
+
             if not isinstance(result.embedding, torch.Tensor):
                 raise ValueError(f"Invalid embeddings file: {file_path.name}")
 

invokeai/backend/model_management/model_manager.py

@@ -249,7 +249,7 @@ from .model_cache import ModelCache, ModelLocker
 from .models import (
     BaseModelType, ModelType, SubModelType,
     ModelError, SchedulerPredictionType, MODEL_CLASSES,
-    ModelConfigBase,
+    ModelConfigBase, ModelNotFoundException,
     )
 
 # We are only starting to number the config file with release 3.
@@ -409,7 +409,7 @@ class ModelManager(object):
         if model_key not in self.models:
             self.scan_models_directory(base_model=base_model, model_type=model_type)
             if model_key not in self.models:
-                raise Exception(f"Model not found - {model_key}")
+                raise ModelNotFoundException(f"Model not found - {model_key}")
 
         model_config = self.models[model_key]
         model_path = self.app_config.root_path / model_config.path
@@ -421,7 +421,7 @@ class ModelManager(object):
 
             else:
                 self.models.pop(model_key, None)
-                raise Exception(f"Model not found - {model_key}")
+                raise ModelNotFoundException(f"Model not found - {model_key}")
 
         # vae/movq override
         # TODO: 

invokeai/backend/model_management/models/__init__.py

@@ -2,7 +2,7 @@ import inspect
 from enum import Enum
 from pydantic import BaseModel
 from typing import Literal, get_origin
-from .base import BaseModelType, ModelType, SubModelType, ModelBase, ModelConfigBase, ModelVariantType, SchedulerPredictionType, ModelError, SilenceWarnings
+from .base import BaseModelType, ModelType, SubModelType, ModelBase, ModelConfigBase, ModelVariantType, SchedulerPredictionType, ModelError, SilenceWarnings, ModelNotFoundException
 from .stable_diffusion import StableDiffusion1Model, StableDiffusion2Model
 from .vae import VaeModel
 from .lora import LoRAModel

invokeai/backend/model_management/models/base.py

@@ -15,6 +15,9 @@ from contextlib import suppress
 from pydantic import BaseModel, Field
 from typing import List, Dict, Optional, Type, Literal, TypeVar, Generic, Callable, Any, Union
 
+class ModelNotFoundException(Exception):
+    pass
+
 class BaseModelType(str, Enum):
     StableDiffusion1 = "sd-1"
     StableDiffusion2 = "sd-2"

invokeai/backend/model_management/models/textual_inversion.py

@@ -8,6 +8,7 @@ from .base import (
     ModelType,
     SubModelType,
     classproperty,
+    ModelNotFoundException,
 )
 # TODO: naming
 from ..lora import TextualInversionModel as TextualInversionModelRaw
@@ -37,8 +38,15 @@ class TextualInversionModel(ModelBase):
         if child_type is not None:
             raise Exception("There is no child models in textual inversion")
 
+        checkpoint_path = self.model_path
+        if os.path.isdir(checkpoint_path):
+            checkpoint_path = os.path.join(checkpoint_path, "learned_embeds.bin")
+
+        if not os.path.exists(checkpoint_path):
+            raise ModelNotFoundException()
+
         model = TextualInversionModelRaw.from_checkpoint(
-            file_path=self.model_path,
+            file_path=checkpoint_path,
             dtype=torch_dtype,
         )