physioex.models.load_embeddings

physioex.models.load_embeddings(model_name, dataset_name, cache_dir=None, verbose=False)

Load cached embeddings, downloading from HuggingFace if needed.

Checks the local cache first. If embeddings are not found locally, downloads them from 4rooms/{model_name}-embeddings on HuggingFace Hub (one dataset repo per model).

Parameters:

• model_name (str) – Model identifier (e.g. "seqsleepnet-phan").

• dataset_name (str) – Dataset name (e.g. "sleepedf").

• cache_dir (str | None) – Override cache root directory.

• verbose (bool) – If True, print embedding metadata and linear probe results (if available).

Returns:

Path to the local embeddings directory containing per-subject subdirectories with embeddings.npy and labels.npy.

Return type:

Path

Example:

from physioex.models import load_embeddings

path = load_embeddings("seqsleepnet-phan", "sleepedf", verbose=True)
# path / "SC4001E0" / "embeddings.npy"  ->  (n_epochs, 128)