invoke-ai/InvokeAI · error · ValueError

Only Gemma2Encoder_Gemma2Encoder_Config models are supported

Error message

Only Gemma2Encoder_Gemma2Encoder_Config models are supported here.

What it means

Gemma2EncoderLoader._load_model is registered only for Gemma2Encoder models and accepts a single config type: Gemma2Encoder_Gemma2Encoder_Config. If the model manager dispatches a config of any other class to this loader (a routing/registration bug or wrong model record), it refuses with this ValueError rather than loading the wrong model.

Source

Thrown at invokeai/backend/model_manager/load/model_loaders/gemma2_encoder.py:84

            out["embed_tokens.weight"] = value
        elif key == "output_norm.weight":
            out["norm.weight"] = value
        else:
            raise ValueError(f"Unmapped Gemma-2 GGUF tensor key '{key}'")
    return out


@ModelLoaderRegistry.register(base=BaseModelType.Any, type=ModelType.Gemma2Encoder, format=ModelFormat.Gemma2Encoder)
class Gemma2EncoderLoader(ModelLoader):
    """Loads a Gemma-2 causal LM directory and exposes its decoder + tokenizer."""

    def _load_model(
        self,
        config: AnyModelConfig,
        submodel_type: Optional[SubModelType] = None,
    ) -> AnyModel:
        if not isinstance(config, Gemma2Encoder_Gemma2Encoder_Config):
            raise ValueError("Only Gemma2Encoder_Gemma2Encoder_Config models are supported here.")

        model_path = Path(config.path)

        match submodel_type:
            case SubModelType.Tokenizer:
                return AutoTokenizer.from_pretrained(model_path, local_files_only=True)
            case SubModelType.TextEncoder:
                target_device = TorchDevice.choose_torch_device()
                model_dtype = TorchDevice.choose_bfloat16_safe_dtype(target_device)
                causal_lm = AutoModelForCausalLM.from_pretrained(
                    model_path,
                    torch_dtype=model_dtype,
                    low_cpu_mem_usage=True,
                    local_files_only=True,
                )
                # PiD only ever uses the decoder block — the transformer stack
                # without the LM head. Upstream calls `.get_decoder()`, but
                # transformers 4.56 returns None for Gemma2, so we reach for

View on GitHub (pinned to 0b6a024f2f)

Solutions

Ensure the model's format is GGUFQuantized so the Gemma2EncoderGGUFLoader handles it instead
Reinstall/re-convert the model so its config is Gemma2Encoder_Gemma2Encoder_Config
Check the registry registration format tags so dispatch selects the correct loader

Example fix

# before
config = Gemma2Encoder_GGUF_Config(...)  # routed to Gemma2EncoderLoader
# after
model_format = ModelFormat.GGUFQuantized  # dispatched to Gemma2EncoderGGUFLoader

Defensive patterns

Strategy: type-guard

Validate before calling

from invokeai.backend.model_manager.config import Gemma2Encoder_Gemma2Encoder_Config
def can_load_with_gemma2_encoder_loader(config):
    return isinstance(config, Gemma2Encoder_Gemma2Encoder_Config)

Type guard

def is_gemma2_encoder_config(config) -> bool:
    return isinstance(config, Gemma2Encoder_Gemma2Encoder_Config)

Try / catch

try:
    model = loader._load_model(config, submodel_type)
except ValueError as e:
    if "Only Gemma2Encoder_Gemma2Encoder_Config" in str(e):
        print(f"Wrong loader for model {config.path}; check its recorded format")
    else:
        raise

Prevention

Ensure model records store the correct ModelFormat so registry dispatch picks the right loader
Never call _load_model directly with configs of a different class
Re-convert models rather than hand-editing config/format metadata

When it happens

Trigger: A model record whose config class is not Gemma2Encoder_Gemma2Encoder_Config is routed to the Gemma2Encoder loader — e.g. a GGUF record handed to the non-GGUF loader because ModelFormat was recorded as Gemma2Encoder instead of GGUFQuantized, or a stale/incorrect models.yaml/DB entry.

Common situations: Model installed with the wrong format metadata so the registry picks the wrong loader; code calling Gemma2EncoderLoader._load_model directly with a foreign config object; refactors renaming the config class without updating model records.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/8952b247fbd8d8c4. Report an issue: GitHub.