invoke-ai/InvokeAI · info · NotAMatchError

missing text_encoder_2/model.safetensors.index.json

Error message

missing text_encoder_2/model.safetensors.index.json

What it means

NotAMatchError raised by T5Encoder_BnBLLMint8_Config.raise_if_doesnt_have_unquantized_config_file during model identification. The bnb LLM.int8 T5 encoder config requires a `text_encoder_2/model.safetensors.index.json` sharded-index file to positively identify the model as a FLUX-style T5 encoder; without it this candidate config declines the match. In InvokeAI's scan pipeline this is a normal probe rejection, not a crash.

Source

Thrown at invokeai/backend/model_manager/configs/t5_encoder.py:60

    def from_model_on_disk(cls, mod: ModelOnDisk, override_fields: dict[str, Any]) -> Self:
        raise_if_not_dir(mod)

        raise_for_override_fields(cls, override_fields)

        expected_config_path = mod.path / "text_encoder_2" / "config.json"
        expected_class_name = "T5EncoderModel"
        raise_for_class_name(expected_config_path, expected_class_name)

        cls.raise_if_doesnt_have_unquantized_config_file(mod)

        return cls(**override_fields)

    @classmethod
    def raise_if_doesnt_have_unquantized_config_file(cls, mod: ModelOnDisk) -> None:
        has_unquantized_config = (mod.path / "text_encoder_2" / "model.safetensors.index.json").exists()

        if not has_unquantized_config:
            raise NotAMatchError("missing text_encoder_2/model.safetensors.index.json")


class T5Encoder_BnBLLMint8_Config(Config_Base):
    """Configuration for T5 Encoder models quantized by bitsandbytes' LLM.int8."""

    base: Literal[BaseModelType.Any] = Field(default=BaseModelType.Any)
    type: Literal[ModelType.T5Encoder] = Field(default=ModelType.T5Encoder)
    format: Literal[ModelFormat.BnbQuantizedLlmInt8b] = Field(default=ModelFormat.BnbQuantizedLlmInt8b)
    cpu_only: bool | None = Field(default=None, description="Whether this model should run on CPU only")

    @classmethod
    def from_model_on_disk(cls, mod: ModelOnDisk, override_fields: dict[str, Any]) -> Self:
        raise_if_not_dir(mod)

        raise_for_override_fields(cls, override_fields)

        expected_config_path = mod.path / "text_encoder_2" / "config.json"
        expected_class_name = "T5EncoderModel"

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Ensure `text_encoder_2/model.safetensors.index.json` exists at the model root (re-download the T5 encoder files from the source diffusers repo)
2. If the T5 weights are a single unsharded model.safetensors, this config legitimately does not apply; let another config match or convert/reshard the model
3. Verify directory layout: model_root/text_encoder_2/ must contain the index json, not a nested subfolder

Example fix

// before
flux-model/
  text_encoder_2/
    model-00001-of-00002.safetensors
// after
flux-model/
  text_encoder_2/
    model.safetensors.index.json
    model-00001-of-00002.safetensors
    model-00002-of-00002.safetensors

from pathlib import Path
def has_bnb_t5_index(model_dir: Path) -> bool:
    return (model_dir / "text_encoder_2" / "model.safetensors.index.json").exists()

try:
    install_model(path)
except NotAMatchError as e:
    if "missing text_encoder_2/model.safetensors.index.json" in str(e):
        logger.warning("T5 encoder is not a sharded llm_int8 layout; trying generic config")
        install_model(path, config_path=GENERIC_T5_CONFIG)

When it happens

Trigger: Calling from_model_on_disk on a model dir where (path/text_encoder_2/model.safetensors.index.json) does not exist — e.g. single-file text_encoder_2 weights, a diffusers layout using model.safetensors (unsharded), or the T5 weights stored elsewhere.

Common situations: Downloading only part of a FLUX repo (skipping sharded T5 files), converting a checkpoint to a single safetensors file, placing text_encoder_2 weights at the wrong nesting level, or manually reorganizing diffusers folders.

Related errors

• directory is not a full FLUX.2 pipeline (no model_index.json
• unrecognized scheduler prediction_type {prediction_type}
• directory looks like a full diffusers pipeline (has model_in
• directory looks like a full diffusers pipeline (has model_in
• missing text_encoder/ subfolder

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/d803fc6f67078df2. Report an issue: GitHub.