invoke-ai/InvokeAI · error · ValueError

No Mistral encoder source provided. Single-file / GGUF trans

Error message

No Mistral encoder source provided. Single-file / GGUF transformers require a separate text encoder. Options:
  1. Set 'Mistral Encoder' to a standalone Mistral Small 3.1 text encoder model
  2. Set 'Mistral Source' to a Diffusers FLUX.2 [dev] model to extract the encoder from

What it means

The FLUX.2 [dev] model loader throws this when a single-file or GGUF transformer is used but no Mistral text encoder source is provided. Such checkpoints lack the text encoder, so InvokeAI needs a standalone Mistral Small 3.1 encoder or a Diffusers FLUX.2 [dev] pipeline to extract the encoder (and tokenizer) from. Raised as ValueError in invoke() when all alternative sources are unset.

Source

Thrown at invokeai/app/invocations/flux2_dev_model_loader.py:158

                "No VAE source provided. Single-file / GGUF transformers require a separate VAE. "
                "Options:\n"
                "  1. Set 'VAE' to a standalone FLUX.2 VAE model\n"
                "  2. Set 'Mistral Source' to a Diffusers FLUX.2 [dev] model to extract the VAE from"
            )

        # Resolve Mistral encoder.
        if self.mistral_encoder_model is not None:
            tokenizer = self.mistral_encoder_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
            text_encoder = self.mistral_encoder_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
        elif main_is_diffusers:
            tokenizer = self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
            text_encoder = self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
        elif self.mistral_source_model is not None:
            self._validate_encoder_source(context, self.mistral_source_model, "Mistral Source")
            tokenizer = self.mistral_source_model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
            text_encoder = self.mistral_source_model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
        else:
            raise ValueError(
                "No Mistral encoder source provided. Single-file / GGUF transformers require a separate "
                "text encoder. Options:\n"
                "  1. Set 'Mistral Encoder' to a standalone Mistral Small 3.1 text encoder model\n"
                "  2. Set 'Mistral Source' to a Diffusers FLUX.2 [dev] model to extract the encoder from"
            )

        return Flux2DevModelLoaderOutput(
            transformer=TransformerField(transformer=transformer, loras=[]),
            mistral_encoder=MistralEncoderField(tokenizer=tokenizer, text_encoder=text_encoder),
            vae=VAEField(vae=vae),
            max_seq_len=self.max_seq_len,
        )

    def _validate_diffusers_format(
        self, context: InvocationContext, model: ModelIdentifierField, model_name: str
    ) -> AnyModelConfig:
        """Validate that a model is a Diffusers-format pipeline and return its config.

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Connect a standalone Mistral Small 3.1 text encoder model to the 'Mistral Encoder' input.
2. Connect a Diffusers FLUX.2 [dev] model to the 'Mistral Source' input to extract the encoder from it.

Example fix

// before
loader = Flux2DevModelLoader(model=gguf_transformer)
// after
loader = Flux2DevModelLoader(model=gguf_transformer, mistral_encoder=mistral_small_3_1_encoder)

# before invoking the loader
if not mistral_encoder_input and not mistral_source_input:
    raise ValueError("Single-file/GGUF transformers need a 'Mistral Encoder' or a Diffusers 'Mistral Source'")

def has_encoder_source(loader) -> bool:
    return loader.mistral_encoder is not None or loader.mistral_source_model is not None

try:
    output = loader.invoke(context)
except ValueError as e:
    if "No Mistral encoder source provided" in str(e):
        loader.mistral_source_model = diffusers_flux2_dev_model
        output = loader.invoke(context)
    else:
        raise

When it happens

Trigger: invoke() runs with a non-Diffusers (single-file/GGUF) main model, the 'Mistral Encoder' input is unset, and self.mistral_source_model is None.

Common situations: Users load a GGUF FLUX.2 transformer for lower VRAM usage but forget that the text encoder must come from elsewhere, leaving the 'Mistral Encoder' and 'Mistral Source' node inputs empty.

Related errors

• No VAE source provided. Single-file / GGUF transformers requ
• The {model_name} model must be a FLUX.2 [dev] pipeline, but
• Expected PreTrainedModel for text encoder, got {type(text_en
• Mistral encoder did not return hidden_states. Ensure output_
• Mistral encoder returned only {num_layers} hidden layer(s),

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/a3234ce33b37b576. Report an issue: GitHub.