8 Commits (test)

Author	SHA1	Message	Date
Michael Nguyen	0354b95af5	restore previous attention cell implemention	2018-09-14 14:13:46 -05:00
Michael Nguyen	9ee48c9286	fixed dropout and prenet bug	2018-09-05 12:38:03 -05:00
Keith Ito	70ba4aaf69	Use vanilla exponential decay for learning rate	2018-03-17 14:24:38 -07:00
Keith Ito	14ee2615eb	Make embedding dimension configurable	2018-03-17 14:22:24 -07:00
Keith Ito	1ff4995eb4	Use location-sensitive attention (hybrid)	2018-03-17 14:08:09 -07:00
Keith Ito	479976b6c5	Simplify text processing, make it easier to adapt to non-English data.	2017-09-04 20:04:59 -07:00
Yunchao He	99518173c0	use clipped_gradients not the original gradients ... To my understanding, it's better to use the `clipped_gradients` not the original `gradients`. Is that right?	2017-07-24 17:32:07 +08:00
Keith Ito	bc706654ad	Initial commit	2017-07-08 13:08:26 -04:00