Fix support for 8b quantized t5 encoders, update exception messages in flux loaders

Author: brandonrising

invokeai/app/invocations/model.py

@@ -143,9 +143,9 @@ def invoke(self, context: InvocationContext) -> ModelIdentifierOutput:
         "format": ModelFormat.T5Encoder,
     },
     "8b_quantized": {
-        "repo": "invokeai/flux_dev::t5_xxl_encoder/optimum_quanto_qfloat8",
+        "repo": "invokeai/flux_schnell::t5_xxl_encoder/optimum_quanto_qfloat8",
         "name": "t5_8b_quantized_encoder",
-        "format": ModelFormat.T5Encoder,
+        "format": ModelFormat.T5Encoder8b,
     },
 }
 

invokeai/backend/model_manager/config.py

@@ -225,6 +225,14 @@ def get_tag() -> Tag:
         return Tag(f"{ModelType.T5Encoder.value}.{ModelFormat.T5Encoder.value}")
 
 
+class T5Encoder8bConfig(T5EncoderConfigBase):
+    format: Literal[ModelFormat.T5Encoder8b] = ModelFormat.T5Encoder8b
+
+    @staticmethod
+    def get_tag() -> Tag:
+        return Tag(f"{ModelType.T5Encoder.value}.{ModelFormat.T5Encoder8b.value}")
+
+
 class LoRALyCORISConfig(LoRAConfigBase):
     """Model config for LoRA/Lycoris models."""
 
@@ -460,6 +468,7 @@ def get_model_discriminator_value(v: Any) -> str:
         Annotated[LoRALyCORISConfig, LoRALyCORISConfig.get_tag()],
         Annotated[LoRADiffusersConfig, LoRADiffusersConfig.get_tag()],
         Annotated[T5EncoderConfig, T5EncoderConfig.get_tag()],
+        Annotated[T5Encoder8bConfig, T5Encoder8bConfig.get_tag()],
         Annotated[TextualInversionFileConfig, TextualInversionFileConfig.get_tag()],
         Annotated[TextualInversionFolderConfig, TextualInversionFolderConfig.get_tag()],
         Annotated[IPAdapterInvokeAIConfig, IPAdapterInvokeAIConfig.get_tag()],

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

@@ -28,12 +28,14 @@
     MainBnbQuantized4bCheckpointConfig,
     MainCheckpointConfig,
     T5EncoderConfig,
+    T5Encoder8bConfig,
     VAECheckpointConfig,
 )
 from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
 from invokeai.backend.model_manager.load.model_loaders.generic_diffusers import GenericDiffusersLoader
 from invokeai.backend.quantization.bnb_nf4 import quantize_model_nf4
 from invokeai.backend.util.silence_warnings import SilenceWarnings
+from invokeai.backend.quantization.fast_quantized_transformers_model import FastQuantizedTransformersModel
 
 app_config = get_config()
 
@@ -82,15 +84,36 @@ def _load_model(
         submodel_type: Optional[SubModelType] = None,
     ) -> AnyModel:
         if not isinstance(config, CLIPEmbedDiffusersConfig):
-            raise Exception("Only Checkpoint Flux models are currently supported.")
+            raise Exception("Only CLIPEmbedDiffusersConfig models are currently supported here.")
 
         match submodel_type:
             case SubModelType.Tokenizer:
                 return CLIPTokenizer.from_pretrained(config.path, max_length=77)
             case SubModelType.TextEncoder:
                 return CLIPTextModel.from_pretrained(config.path)
 
-        raise Exception("Only Checkpoint Flux models are currently supported.")
+        raise Exception("Only Tokenizer and TextEncoder submodels are currently supported.")
+
+
+@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.T5Encoder, format=ModelFormat.T5Encoder8b)
+class T5Encoder8bCheckpointModel(GenericDiffusersLoader):
+    """Class to load main models."""
+
+    def _load_model(
+        self,
+        config: AnyModelConfig,
+        submodel_type: Optional[SubModelType] = None,
+    ) -> AnyModel:
+        if not isinstance(config, T5Encoder8bConfig):
+            raise Exception("Only T5Encoder8bConfig models are currently supported here.")
+
+        match submodel_type:
+            case SubModelType.Tokenizer2:
+                return T5Tokenizer.from_pretrained(Path(config.path) / "tokenizer_2", max_length=512)
+            case SubModelType.TextEncoder2:
+                return FastQuantizedTransformersModel.from_pretrained(Path(config.path) / "text_encoder_2")
+
+        raise Exception("Only Tokenizer and TextEncoder submodels are currently supported.")
 
 
 @ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.T5Encoder, format=ModelFormat.T5Encoder)
@@ -103,7 +126,7 @@ def _load_model(
         submodel_type: Optional[SubModelType] = None,
     ) -> AnyModel:
         if not isinstance(config, T5EncoderConfig):
-            raise Exception("Only Checkpoint Flux models are currently supported.")
+            raise Exception("Only T5EncoderConfig models are currently supported here.")
 
         match submodel_type:
             case SubModelType.Tokenizer2:
@@ -113,7 +136,7 @@ def _load_model(
                     Path(config.path) / "text_encoder_2"
                 )  # TODO: Fix hf subfolder install
 
-        raise Exception("Only Checkpoint Flux models are currently supported.")
+        raise Exception("Only Tokenizer and TextEncoder submodels are currently supported.")
 
 
 @ModelLoaderRegistry.register(base=BaseModelType.Flux, type=ModelType.Main, format=ModelFormat.Checkpoint)
@@ -126,7 +149,7 @@ def _load_model(
         submodel_type: Optional[SubModelType] = None,
     ) -> AnyModel:
         if not isinstance(config, CheckpointConfigBase):
-            raise Exception("Only Checkpoint Flux models are currently supported.")
+            raise Exception("Only CheckpointConfigBase models are currently supported here.")
         legacy_config_path = app_config.legacy_conf_path / config.config_path
         config_path = legacy_config_path.as_posix()
         with open(config_path, "r") as stream:
@@ -139,7 +162,7 @@ def _load_model(
             case SubModelType.Transformer:
                 return self._load_from_singlefile(config, flux_conf)
 
-        raise Exception("Only Checkpoint Flux models are currently supported.")
+        raise Exception("Only Transformer submodels are currently supported.")
 
     def _load_from_singlefile(
         self,
@@ -171,7 +194,7 @@ def _load_model(
         submodel_type: Optional[SubModelType] = None,
     ) -> AnyModel:
         if not isinstance(config, CheckpointConfigBase):
-            raise Exception("Only Checkpoint Flux models are currently supported.")
+            raise Exception("Only CheckpointConfigBase models are currently supported here.")
         legacy_config_path = app_config.legacy_conf_path / config.config_path
         config_path = legacy_config_path.as_posix()
         with open(config_path, "r") as stream:
@@ -184,7 +207,7 @@ def _load_model(
             case SubModelType.Transformer:
                 return self._load_from_singlefile(config, flux_conf)
 
-        raise Exception("Only Checkpoint Flux models are currently supported.")
+        raise Exception("Only Transformer submodels are currently supported.")
 
     def _load_from_singlefile(
         self,