Commits

Commits on Jun 22, 2017

Use bucketing sampler while continue_from
EgorLakomkin
committedJun 22, 2017

Commits on Jun 16, 2017

save wer/cer and avg_loss during training
EgorLakomkin
committedJun 16, 2017

Commits on Jun 14, 2017

+info about SpectrogramDatasetWithLength
EgorLakomkin
committedJun 14, 2017
deleting loss and out references to prevent OOMs, fixed duration checking for files with space in the path
EgorLakomkin
committedJun 14, 2017

Commits on Jun 11, 2017

+bucketing
EgorLakomkin
committedJun 11, 2017

Commits on May 12, 2017

filtering by min/max duration individually
EgorLakomkin
committedMay 12, 2017

Commits on May 6, 2017

Commits on Apr 28, 2017

minor corrections
EgorLakomkin
committedApr 28, 2017
added librispeech corpus
EgorLakomkin
committedApr 28, 2017
added LibriSpeech corpus
EgorLakomkin
committedApr 28, 2017

Commits on Apr 27, 2017

passing only rnn class to the deepspeech module
EgorLakomkin
committedApr 27, 2017

Commits on Apr 26, 2017

added support of simple_rnn, gru and lstm for the deepspeech module
EgorLakomkin
committedApr 26, 2017

Commits on Apr 22, 2017

handle flac samples, filter unused characters
EgorLakomkin
committedApr 22, 2017

Commits on Apr 21, 2017

downloading 16kHz samples from VoxForge
EgorLakomkin
committedApr 21, 2017

Commits on Apr 20, 2017

added TEDLIUM_release2 corpus
EgorLakomkin
committedApr 20, 2017