invoke-ai/InvokeAI · error · NotAMatchError

unable to load config file: {config_path_nested} does not ex

Error message

unable to load config file: {config_path_nested} does not exist

What it means

The standalone Qwen3-VL encoder config expects a config.json either in a text_encoder/ subfolder or at the directory root. When neither exists it cannot identify the architecture and throws this NotAMatchError (the message names the nested path it probed first). Without config.json InvokeAI cannot confirm the directory is a Qwen3VLModel.

Source

Thrown at invokeai/backend/model_manager/configs/qwen3_vl_encoder.py:125

        # Exclude full pipeline models - these should be matched as main models, not just encoders.
        model_index_path = mod.path / "model_index.json"
        transformer_path = mod.path / "transformer"
        if model_index_path.exists() or transformer_path.exists():
            raise NotAMatchError(
                "directory looks like a full diffusers pipeline (has model_index.json or transformer folder), "
                "not a standalone Qwen3-VL encoder"
            )

        # Support both a nested text_encoder/config.json and a standalone config.json at the root.
        config_path_nested = mod.path / "text_encoder" / "config.json"
        config_path_direct = mod.path / "config.json"

        if config_path_nested.exists():
            expected_config_path = config_path_nested
        elif config_path_direct.exists():
            expected_config_path = config_path_direct
        else:
            raise NotAMatchError(f"unable to load config file: {config_path_nested} does not exist")

        # Qwen3-VL uses the Qwen3VLModel / Qwen3VLForConditionalGeneration architecture.
        raise_for_class_name(
            expected_config_path,
            {
                "Qwen3VLModel",
                "Qwen3VLForConditionalGeneration",
            },
        )
        _validate_krea2_qwen3_vl_config(expected_config_path)

        if config_path_nested.exists():
            weights_path = mod.path / "text_encoder"
            tokenizer_path = mod.path / "tokenizer"
        else:
            weights_path = mod.path
            tokenizer_path = mod.path

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Add the missing config.json: place it at the folder root, or inside a text_encoder/ subfolder alongside the weights.
2. Re-download the model directory from HuggingFace completely, ensuring config.json is included (check ignore patterns like *.json filters).
3. If the config lives in another subfolder, restructure so it is at mod.path/config.json or mod.path/text_encoder/config.json.
4. Verify the config's architectures field lists Qwen3VLModel or Qwen3VLForConditionalGeneration once the file is present.

Example fix

// before
models/qwen3vl-encoder/
  model-00001-of-00002.safetensors
  (no config.json) -> NotAMatchError
// after
models/qwen3vl-encoder/
  config.json
  model-00001-of-00002.safetensors
  model-00002-of-00002.safetensors

from pathlib import Path

def has_qwen3vl_config(model_dir: str) -> bool:
    p = Path(model_dir)
    return (p / "text_encoder" / "config.json").exists() or (p / "config.json").exists()

def config_json_present(p) -> bool:
    from pathlib import Path
    p = Path(p)
    return (p / "config.json").is_file() or (p / "text_encoder" / "config.json").is_file()

try:
    cfg = Qwen3VLEncoder_Qwen3VLEncoder_Config.from_model_on_disk(mod, {})
except NotAMatchError as e:
    if "does not exist" in str(e) and "config.json" in str(e):
        logger.warning("Missing config.json in %s; re-download the model directory", mod.path)

When it happens

Trigger: from_model_on_disk is invoked on a directory where both mod.path/text_encoder/config.json and mod.path/config.json are missing - e.g. a folder with only weight shards, or a ComfyUI-style folder where config lives under a differently named subfolder.

Common situations: Downloading only the weights from HuggingFace (config.json skipped by .gitignore or partial download), copying weights without their config, renaming text_encoder to another name, or pointing the import at a nested folder one level too deep/shallow.

Related errors

• standalone Qwen3-VL encoder directory does not contain model
• directory does not contain Gemma2 tokenizer files (tokenizer
• missing ip_adapter.bin weights file
• missing image_encoder.txt metadata file
• unrecognized token vector length {token_vector_length}

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/30677ff3da7bc639. Report an issue: GitHub.