espnet.utils.cli_readers.file_reader_helper

source

espnet.utils.cli_readers.file_reader_helper(rspecifier: str, filetype: str = 'mat', return_shape: bool = False, segments: str | None = None)

Read uttid and array in kaldi style

This function might be a bit confusing as “ark” is used for HDF5 to imitate “kaldi-rspecifier”.

• Parameters:
  • rspecifier – Give as “ark:feats.ark” or “scp:feats.scp”
  • filetype – “mat” is kaldi-martix, “hdf5”: HDF5
  • return_shape – Return the shape of the matrix, instead of the matrix. This can reduce IO cost for HDF5.
• Return type:Generator[Tuple[str, np.ndarray], None, None]

Examples

Read from kaldi-matrix ark file:

>>> for u, array in file_reader_helper('ark:feats.ark', 'mat'):
...     array

Read from HDF5 file:

>>> for u, array in file_reader_helper('ark:feats.h5', 'hdf5'):
...     array