taylorlu / ghostvlad-speaker

An tensorflow implementation of ghostvlad for speaker recognition
14 stars 9 forks source link

Not keeping running after 0th epoch ends. #2

Closed gogyzzz closed 5 years ago

gogyzzz commented 5 years ago

image

num of workers: 24

After iteration of 426x, all processes in iterator() function seems to be stopped. image

But the next epoch won't start.

taylorlu commented 5 years ago

Sorry, there seems has a bug in audio_producer.py when finish an epoch. you can renew an instance when finished an epoch in train.py

    producer = audio_producer.AudioProducer(argv['json_path'], argv['batch_size'],
                                            sample_rate=argv['sample_rate'],
                                            min_duration=argv['min_duration'],
                                            max_duration=argv['max_duration'])
gogyzzz commented 5 years ago

Did you mean putting the code after run_epoch() ?

I found the function works well when I cut the length of number of batches. image

gogyzzz commented 5 years ago

I used the trick like below, and it works. image