8 Commits

Author	SHA1	Message	Date
nttstar	82fbd60164	support weight matrix parall training on single machine	2019-02-09 20:55:26 +08:00
nttstar	849f35e3ab	add densenet dropout config	2019-01-21 19:46:36 +08:00
nttstar	f012481dfb	fix	2019-01-21 19:12:45 +08:00
nttstar	1813610699	add densenet	2019-01-21 19:08:43 +08:00
nttstar	a661e00ad1	add mnasnet; add save embedding only option	2019-01-18 13:23:34 +08:00
nttstar	55a25c7620	tiny	2019-01-17 16:56:08 +08:00
nttstar	9c8d765f9c	move more to config	2019-01-17 16:42:39 +08:00
nttstar	70f7864a35	support both softmax and triplet loss in train.py	2019-01-15 22:13:10 +08:00