Use tf.contrib.cudnn_rnn or at least tf.contrib.rnn.BlockLSTMCell

mozilla / DeepSpeech

DeepSpeech is an open source embedded (offline, on-device) speech-to-text engine which can run in real time on devices ranging from a Raspberry Pi 4 to high power GPU servers.

Mozilla Public License 2.0

25.29k stars 3.96k forks source link

Use tf.contrib.cudnn_rnn or at least tf.contrib.rnn.BlockLSTMCell #1110

Closed ekelsen closed 6 years ago

ekelsen commented 6 years ago

https://github.com/mozilla/DeepSpeech/blob/cbc868346bd0a0c0e9745a978f5818364a33084c/DeepSpeech.py#L424

BasicLSTMCell is a reference implementation, it is quite slow and memory inefficient. It would be much better to use https://www.tensorflow.org/api_docs/python/tf/contrib/cudnn_rnn (literally an order of magnitude faster) or at least https://www.tensorflow.org/api_docs/python/tf/contrib/rnn/LSTMBlockCell .

kdavis-mozilla commented 6 years ago

Thanks for the heads up. But this is a duplicate of issue #885

ekelsen commented 6 years ago

Using LSTMBlockCell is literally a regex that might be worth doing if implementing cudnn is taking months...

lock[bot] commented 5 years ago

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.