record model_variant in t2i and clip_vision configs (#5989)

- Move base of t2i and clip_vision config models to DiffusersBase, which contains
  a field to record the model variant (e.g. "fp16")
- This restore the ability to load fp16 t2i and clip_vision models
- Also add defensive coding to load the vanilla model when the fp16 model
  has been replaced (or more likely, user's preferences changed since installation)

Co-authored-by: Lincoln Stein <lstein@gmail.com>

invokeai/backend/model_manager/config.py

@@ -331,7 +331,7 @@ class IPAdapterConfig(ModelConfigBase):
         return Tag(f"{ModelType.IPAdapter.value}.{ModelFormat.InvokeAI.value}")
 
 
-class CLIPVisionDiffusersConfig(ModelConfigBase):
+class CLIPVisionDiffusersConfig(DiffusersConfigBase):
     """Model config for CLIPVision."""
 
     type: Literal[ModelType.CLIPVision] = ModelType.CLIPVision
@@ -342,7 +342,7 @@ class CLIPVisionDiffusersConfig(ModelConfigBase):
         return Tag(f"{ModelType.CLIPVision.value}.{ModelFormat.Diffusers.value}")
 
 
-class T2IAdapterConfig(ModelConfigBase, ControlAdapterConfigBase):
+class T2IAdapterConfig(DiffusersConfigBase, ControlAdapterConfigBase):
     """Model config for T2I."""
 
     type: Literal[ModelType.T2IAdapter] = ModelType.T2IAdapter

invokeai/backend/model_manager/load/model_loaders/generic_diffusers.py

@@ -36,7 +36,15 @@ class GenericDiffusersLoader(ModelLoader):
         if submodel_type is not None:
             raise Exception(f"There are no submodels in models of type {model_class}")
         variant = model_variant.value if model_variant else None
-        result: AnyModel = model_class.from_pretrained(model_path, torch_dtype=self._torch_dtype, variant=variant)  # type: ignore
+        try:
+            result: AnyModel = model_class.from_pretrained(model_path, torch_dtype=self._torch_dtype, variant=variant)
+        except OSError as e:
+            if variant and "no file named" in str(
+                e
+            ):  # try without the variant, just in case user's preferences changed
+                result = model_class.from_pretrained(model_path, torch_dtype=self._torch_dtype)
+            else:
+                raise e
         return result
 
     # TO DO: Add exception handling
@@ -63,7 +71,7 @@ class GenericDiffusersLoader(ModelLoader):
                     assert class_name is not None
                     result = self._hf_definition_to_type(module="transformers", class_name=class_name[0])
                 if not class_name:
-                    raise InvalidModelConfigException("Unable to decifer Load Class based on given config.json")
+                    raise InvalidModelConfigException("Unable to decipher Load Class based on given config.json")
             except KeyError as e:
                 raise InvalidModelConfigException("An expected config.json file is missing from this model.") from e
         assert result is not None

invokeai/backend/model_manager/load/model_loaders/stable_diffusion.py

@@ -44,11 +44,20 @@ class StableDiffusionDiffusersModel(GenericDiffusersLoader):
         load_class = self.get_hf_load_class(model_path, submodel_type)
         variant = model_variant.value if model_variant else None
         model_path = model_path / submodel_type.value
-        result: AnyModel = load_class.from_pretrained(
-            model_path,
-            torch_dtype=self._torch_dtype,
-            variant=variant,
-        )  # type: ignore
+        try:
+            result: AnyModel = load_class.from_pretrained(
+                model_path,
+                torch_dtype=self._torch_dtype,
+                variant=variant,
+            )
+        except OSError as e:
+            if variant and "no file named" in str(
+                e
+            ):  # try without the variant, just in case user's preferences changed
+                result = load_class.from_pretrained(model_path, torch_dtype=self._torch_dtype)
+            else:
+                raise e
+
         return result
 
     def _needs_conversion(self, config: AnyModelConfig, model_path: Path, dest_path: Path) -> bool: