mlx.core.load

load(file: file | str | Path, /, format: str | None = None, return_metadata: bool = False, *, stream: None | Stream | Device = None) → array | dict[str, array]

Load array(s) from a binary file.

The supported formats are .npy, .npz, .safetensors, and .gguf.

Parameters:

• file (file, str, Path) – File in which the array is saved.

• format (str, optional) – Format of the file. If None, the format is inferred from the file extension. Supported formats: npy, npz, and safetensors. Default: None.

• return_metadata (bool, optional) – Load the metadata for formats which support matadata. The metadata will be returned as an additional dictionary. Default: False.

Returns:

A single array if loading from a .npy file or a dict mapping names to arrays if loading from a .npz or .safetensors file. If return_metadata is True an additional dictionary of metadata will be returned.

Return type:

array or dict

Warning

When loading unsupported quantization formats from GGUF, tensors will automatically cast to mx.float16