ValueError: frames must be specified for non-seekable files

[yygg678]

Preprocessing voxceleb2
VoxCeleb2: Preprocessing data for 5994 speakers.
VoxCeleb2: 0%| | 0/5994 [00:00<?, ?speakers/s]Traceback (most recent call last):
File "encoder_preprocess.py", line 61, in
preprocess_funcdataset
File "/NASdata/yangyg/Real-Time-Voice-Cloning/encoder/preprocess.py", line 175, in preprocess_voxceleb2
skip_existing, logger)
File "/NASdata/yangyg/Real-Time-Voice-Cloning/encoder/preprocess.py", line 116, in _preprocess_speaker_dirs
unit="speakers"))
File "/home/yyg/anaconda3/envs/RealTime/lib/python3.6/site-packages/tqdm/_tqdm.py", line 1017, in iter
for obj in iterable:
File "/home/yyg/anaconda3/envs/RealTime/lib/python3.6/site-packages/multiprocess/pool.py", line 735, in next
raise value
File "/home/yyg/anaconda3/envs/RealTime/lib/python3.6/site-packages/multiprocess/pool.py", line 119, in worker
result = (True, func(*args, **kwds))
File "/NASdata/yangyg/Real-Time-Voice-Cloning/encoder/preprocess.py", line 97, in preprocess_speaker
wav = audio.preprocess_wav(in_fpath)
File "/NASdata/yangyg/Real-Time-Voice-Cloning/encoder/audio.py", line 28, in preprocess_wav
wav, source_sr = librosa.load(fpath_or_wav, sr=None)
File "/home/yyg/anaconda3/envs/RealTime/lib/python3.6/site-packages/librosa/core/audio.py", line 138, in load
y = sf_desc.read(frames=frame_duration, dtype=dtype, always_2d=False).T
File "/home/yyg/anaconda3/envs/RealTime/lib/python3.6/site-packages/soundfile.py", line 860, in read
frames = self._check_frames(frames, fill_value)
File "/home/yyg/anaconda3/envs/RealTime/lib/python3.6/site-packages/soundfile.py", line 1268, in _check_frames
raise ValueError("frames must be specified for non-seekable files")
ValueError: frames must be specified for non-seekable files

now ,what can I do to solve it. thanks.

[sberryman]

@yyggithub I ran into that as well and I found two ways around it.

1. You can change the thread pool size to 1 (from 8) https://github.com/CorentinJ/Real-Time-Voice-Cloning/blob/master/encoder/preprocess.py#L114 Honestly I'm not sure why this worked.
2. You can convert every m4v file to wav. This was the route I went, just wrote a simple batch script and used ffmpeg (probably faster ways) to do the conversion.

[yygg678]

thank you for your answer,I have solve the problem.but I solve it in a different way,just only change librosa version to 0.5.0 (from 0.7.0) I think the mainly problem is thread pool size.