invoke-ai/InvokeAI · error · TypeError

Expected QwenVLEncoder_Diffusers_Config, got {type(config)._

Error message

Expected QwenVLEncoder_Diffusers_Config, got {type(config).__name__}.

What it means

The diffusers-format Qwen VL encoder loader only accepts QwenVLEncoder_Diffusers_Config. If _load_model receives any other config type (e.g. a checkpoint config for the same model type), it raises this TypeError naming the actual type, since the diffusers loader uses from_pretrained-style loading that only applies to diffusers-format model folders.

Source

Thrown at invokeai/backend/model_manager/load/model_loaders/qwen_image.py:302

        new_sd_size = sum(t.nelement() * t.element_size() for t in sd.values())
        self._ram_cache.make_room(new_sd_size)

        model.load_state_dict(sd, strict=False, assign=True)
        return model


@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.QwenVLEncoder, format=ModelFormat.QwenVLEncoder)
class QwenVLEncoderLoader(ModelLoader):
    """Loads a standalone Qwen2.5-VL encoder (text_encoder/ + tokenizer/ + processor/)."""

    def _load_model(
        self,
        config: AnyModelConfig,
        submodel_type: Optional[SubModelType] = None,
    ) -> AnyModel:
        if not isinstance(config, QwenVLEncoder_Diffusers_Config):
            raise TypeError(f"Expected QwenVLEncoder_Diffusers_Config, got {type(config).__name__}.")

        from transformers import AutoTokenizer, Qwen2_5_VLForConditionalGeneration

        model_path = Path(config.path)

        target_device = TorchDevice.choose_torch_device()
        model_dtype = TorchDevice.choose_bfloat16_safe_dtype(target_device)

        match submodel_type:
            case SubModelType.Tokenizer:
                tokenizer_path = model_path / "tokenizer"
                return AutoTokenizer.from_pretrained(str(tokenizer_path), local_files_only=True)
            case SubModelType.TextEncoder:
                encoder_path = model_path / "text_encoder"
                return Qwen2_5_VLForConditionalGeneration.from_pretrained(
                    str(encoder_path),
                    torch_dtype=model_dtype,
                    low_cpu_mem_usage=True,

View on GitHub (pinned to 0b6a024f2f)

Solutions

Ensure the model is registered as diffusers format so QwenVLEncoder_Diffusers_Config is produced and routed here.
If the model is a single-file checkpoint, let it route to the checkpoint-registered QwenVLEncoder loader instead.
Correct the format field on the model record and re-scan.

Example fix

// before
config = QwenVLEncoder_Checkpoint_Config(path="encoder.safetensors")
enc = diffusers_loader._load_model(config, SubModelType.TextEncoder)
// after
config = QwenVLEncoder_Diffusers_Config(path="encoder_dir/")
enc = diffusers_loader._load_model(config, SubModelType.TextEncoder)

Defensive patterns

Strategy: type-guard

Validate before calling

from invokeai.backend.model_manager.config import QwenVLEncoder_Diffusers_Config
if not isinstance(config, QwenVLEncoder_Diffusers_Config):
    raise TypeError(f"Diffusers QwenVL loader needs QwenVLEncoder_Diffusers_Config, got {type(config).__name__}")

Type guard

def is_diffusers_qwenvl_config(config: AnyModelConfig) -> bool:
    return isinstance(config, QwenVLEncoder_Diffusers_Config)

Try / catch

try:
    enc = loader._load_model(config, SubModelType.TextEncoder)
except TypeError as e:
    if "QwenVLEncoder_Diffusers_Config" in str(e):
        enc = checkpoint_qwenvl_loader._load_model(config, SubModelType.TextEncoder)  # route to the other loader
    else:
        raise

Prevention

Verify the model's format field (diffusers vs checkpoint) matches the on-disk layout
Use separate loader instances per format
Re-scan the models directory after moving encoder files

When it happens

Trigger: A Qwen VL text encoder model registered with ModelFormat.Checkpoint but resolved to the diffusers-format loader, or vice versa: checkpoint config object passed to the diffusers loader's _load_model.

Common situations: Model imported from a single-file .safetensors but the diffusers registry entry was chosen; model record's format field wrong after manual edits; mixed encoder setups.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/83fe0eebd4e9ecd4. Report an issue: GitHub.