Skip to content

get_model_package_contents

inference_models.developer_tools.get_model_package_contents 

get_model_package_contents(model_package_dir, elements)

Get absolute paths to files within a model package directory.

This function is primarily used when implementing custom models that need to load files from a model package. It validates that all required files exist and returns their absolute paths.

Parameters:

  • model_package_dir

    (str) –

    Absolute path to the model package directory.

  • elements

    (List[str]) –

    List of file names (relative to package directory) to retrieve.

Returns:

  • Dict[str, str] –

    Dictionary mapping each element name to its absolute path.

Raises:

  • CorruptedModelPackageError –

    If any of the requested elements don't exist in the package directory.

Examples:

Load model weights and config from a package:

>>> from inference_models.developer_tools import get_model_package_contents
>>>
>>> package_contents = get_model_package_contents(
...     model_package_dir="/path/to/model/package",
...     elements=["weights.pt", "config.json", "class_names.txt"]
... )
>>>
>>> # Access file paths
>>> weights_path = package_contents["weights.pt"]
>>> config_path = package_contents["config.json"]
>>> class_names_path = package_contents["class_names.txt"]

Use in custom model implementation:

>>> from inference_models import ObjectDetectionModel
>>> from inference_models.developer_tools import get_model_package_contents
>>> import torch
>>>
>>> class MyDetector(ObjectDetectionModel):
...     @classmethod
...     def from_pretrained(cls, model_name_or_path: str, **kwargs):
...         # Get required files from package
...         contents = get_model_package_contents(
...             model_package_dir=model_name_or_path,
...             elements=["model.onnx", "config.json"]
...         )
...
...         # Load model using the paths
...         model = load_onnx_model(contents["model.onnx"])
...         config = load_config(contents["config.json"])
...
...         return cls(model=model, config=config)
See Also
  • Custom model implementation guide: https://inference.roboflow.com/inference_models/how-to/local-packages/