invoke-ai/InvokeAI · error · ValueError

To extract the VAE and Qwen3-VL encoder, the {model_name} mo

Error message

To extract the VAE and Qwen3-VL encoder, the {model_name} model must be in Diffusers format. The selected model '{config.name}' is in {config.format.value} format — provide a standalone VAE and Qwen3-VL Encoder instead.

What it means

When no standalone VAE or Qwen3-VL encoder is provided, Krea2ModelLoader extracts them as submodels of the main model, which only works if the main model is stored in Diffusers format. Single-file/checkpoint formats cannot yield these submodels, so _validate_diffusers_format raises this ValueError.

Source

Thrown at invokeai/app/invocations/krea2_model_loader.py:124

            text_encoder = self.qwen3_vl_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
        else:
            self._validate_diffusers_format(context, self.model, "Krea-2")
            tokenizer = self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
            text_encoder = self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder})

        return Krea2ModelLoaderOutput(
            transformer=TransformerField(transformer=transformer, loras=[]),
            qwen3_vl_encoder=Qwen3VLEncoderField(tokenizer=tokenizer, text_encoder=text_encoder, loras=[]),
            vae=VAEField(vae=vae),
        )

    def _validate_diffusers_format(
        self, context: InvocationContext, model: ModelIdentifierField, model_name: str
    ) -> None:
        """Validate that a model is in Diffusers format (required to extract VAE / encoder submodels)."""
        config = context.models.get_config(model)
        if config.format != ModelFormat.Diffusers:
            raise ValueError(
                f"To extract the VAE and Qwen3-VL encoder, the {model_name} model must be in Diffusers format. "
                f"The selected model '{config.name}' is in {config.format.value} format — provide a standalone "
                "VAE and Qwen3-VL Encoder instead."
            )

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Provide standalone VAE and Qwen3-VL encoder models in the vae_model and qwen3_vl_encoder_model fields.
2. Convert the main model to Diffusers format (re-import from a full Diffusers repo or use the model manager's convert action) and keep submodel extraction.
3. Download the Diffusers-format variant of the Krea-2 model and install it in place of the checkpoint.

Example fix

// before
Krea2ModelLoader(model=krea2_single_file_checkpoint)  # no vae/encoder given
// after
Krea2ModelLoader(model=krea2_single_file_checkpoint, vae_model=qwen_image_vae, qwen3_vl_encoder_model=qwen3_vl_encoder)

main_cfg = context.models.get_config(model_field)
needs_extraction = vae_field is None or encoder_field is None
if needs_extraction and main_cfg.format is not ModelFormat.Diffusers:
    raise ValueError(f"{main_cfg.name} must be Diffusers format to extract VAE/encoder")

from invokeai.backend.model_manager.config import ModelFormat

def supports_submodel_extraction(cfg) -> bool:
    return cfg.format is ModelFormat.Diffusers

try:
    output = loader.invoke(context)
except ValueError as e:
    if "must be in Diffusers format" in str(e):
        output = loader.model_copy(update={
            "vae_model": default_qwen_vae,
            "qwen3_vl_encoder_model": default_qwen3vl_encoder,
        }).invoke(context)
    else:
        raise

When it happens

Trigger: Calling invoke() (via _validate_diffusers_format) with vae_model unset and/or qwen3_vl_encoder_model unset while the main model config.format is not ModelFormat.Diffusers (e.g. a single-file safetensors/checkpoint Krea-2 model).

Common situations: Using a single-file Krea-2 checkpoint download without converting to Diffusers layout; users who downloaded compact checkpoint versions instead of full Diffusers repos; older installed models predating Diffusers-format import.

Related errors

• The {model_name} model must be a Diffusers format model. The
• The {model_name} model must be a Diffusers-style FLUX.2 pipe
• CheckpointConfigBase is not implemented for Qwen Image Edit
• LoRA '{lora.lora.key}' has conflicting weights on the transf
• Model '{main_config.name}' is not a Krea-2 main model. Selec

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/5187245e25e6196c. Report an issue: GitHub.