invoke-ai/InvokeAI · error · Exception

Model not found: {model_path}

Error message

Model not found: {model_path}

What it means

OnnxRuntimeModel.from_pretrained() accepts a model_id that is either a Hub repo id (downloaded first) or a local path. If the resolved model_path is not an existing file (os.path.isfile fails), it raises this Exception — the ONNX model file itself is missing.

Source

Thrown at invokeai/backend/onnx/onnx_runtime.py:220

        file_name: Optional[str] = None,
        provider: Optional[str] = None,
        sess_options: Optional["SessionOptions"] = None,
        **kwargs: Any,
    ) -> Any:  # fixme
        file_name = file_name or ONNX_WEIGHTS_NAME

        if os.path.isdir(model_id):
            model_path = model_id
            if subfolder is not None:
                model_path = os.path.join(model_path, subfolder)
            model_path = os.path.join(model_path, file_name)

        else:
            model_path = model_id

        # load model from local directory
        if not os.path.isfile(model_path):
            raise Exception(f"Model not found: {model_path}")

        # TODO: session options
        return cls(str(model_path), provider=provider)

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Point model_id at the actual .onnx file (e.g. /path/to/model/model.onnx), not the containing directory.
2. Verify the file exists: ls the directory and confirm the ONNX artifact downloaded completely.
3. If using a Hub id, ensure network access/cache (HF_HOME) so the download succeeds before load.
4. Re-export or re-download the ONNX model if the artifact is missing.

Example fix

// before
pipe = OnnxStableDiffusionPipeline.from_pretrained('/models/sdxl-onnx/')
// after (point at the onnx file expected by the loader, or the repo root containing it)
pipe = OnnxStableDiffusionPipeline.from_pretrained('/models/sdxl-onnx', variant='fp16')

import os
model_path = resolve(model_id)
if not os.path.isfile(model_path):
    raise FileNotFoundError(f'Provide the path to the .onnx file; {model_path} is missing')

from pathlib import Path
def onnx_file_ready(p) -> bool:
    path = Path(p)
    return path.is_file() and path.suffix.lower() == '.onnx'

try:
    model = OnnxRuntimeModel.from_pretrained(model_id, provider=provider)
except Exception as e:
    if str(e).startswith('Model not found:'):
        missing = str(e).split(': ', 1)[1]
        print(f'{missing} does not exist — check the path or re-download the ONNX model')
    else:
        raise

When it happens

Trigger: Passing a directory that contains no ONNX model file, a path to a non-existent location, or a repo id whose download did not produce the expected .onnx file, so `not os.path.isfile(model_path)` triggers.

Common situations: Pointing model_id at the model folder instead of the .onnx file inside it; incomplete/cancelled downloads leaving empty dirs; renamed or moved model directories; expecting from_pretrained to fetch from the Hub while offline with no cached copy.

Related errors

• ImageFileNotFoundException
• Backslashes not allowed in subfolder path
• Absolute paths not allowed in subfolder path
• Empty path segments not allowed in subfolder path
• Source image does not exist: {move.old_path}

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/a2a623b515118c3d. Report an issue: GitHub.