def load_scp(fname, endian='<', separator=None, as_bytes=False, segments=None):
"""Lazy loader for kaldi scp file.
Args:
fname (str or file(text mode)):
endian (str):
separator (str):
as_bytes (bool): Read as raw bytes string
segments (str): The path of segments
"""
assert endian in ('<', '>'), endian
if segments is None:
load_func = partial(load_mat, endian=endian, as_bytes=as_bytes)
loader = LazyLoader(load_func)
with open_or_fd(fname, 'r') as fd:
for line in fd:
seps = line.split(separator, 1)
if len(seps) != 2:
raise ValueError(
'Invalid line is found:\n> {}'.format(line))
token, arkname = seps
loader[token] = arkname.rstrip()
return loader
else:
return SegmentsExtractor(fname, separator=separator, segments=segments)
def load_scp(fname, endian="<", separator=None, segments=None, max_cache_fd=0):
"""Lazy loader for kaldi scp file.
Args:
fname (str or file(text mode)):
endian (str):
separator (str):
segments (str): The path of segments
"""
assert endian in ("<", ">"), endian
if max_cache_fd != 0:
if segments is not None:
raise ValueError("max_cache_fd is not supported for segments mode")
d = LimitedSizeDict(max_cache_fd)
else:
d = None
if segments is None:
load_func = partial(load_mat, endian=endian, fd_dict=d)
loader = LazyLoader(load_func)
with open_like_kaldi(fname, "r") as fd:
for line in fd:
seps = line.split(separator, 1)
if len(seps) != 2:
raise ValueError(
"Invalid line is found:\n> {}".format(line))
token, arkname = seps
loader[token] = arkname.rstrip()
return loader
else:
return SegmentsExtractor(fname, separator=separator, segments=segments)