2439 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Hieu Hoang	2b134fbf49	abort if cuda error	2018-02-02 22:05:05 +00:00
Hieu Hoang	c01867dfae	copy data to device	2018-01-25 14:58:27 +00:00
Hieu Hoang	4984e7fcac	delete strides. Not needed	2018-01-25 14:31:59 +00:00
Hieu Hoang	e1e90ccaf6	move matrix class to its own file	2018-01-25 14:22:40 +00:00
Hieu Hoang	d59f013b49	beef up matrix class	2018-01-25 14:14:13 +00:00
Hieu Hoang	1dec5d658e	matrix class	2018-01-25 12:32:50 +00:00
Hieu Hoang	022db251b8	runs with dummy kernel	2018-01-25 12:16:36 +00:00
Hieu Hoang	23e755a22d	compiles	2018-01-25 11:57:43 +00:00
Hieu Hoang	1554e3884a	opencl test	2018-01-25 11:53:46 +00:00
Hieu Hoang	e083655133	opencl test	2018-01-25 11:49:53 +00:00
Hieu Hoang	cf4e6726d3	add arg --use-fused-softmax	2018-01-22 17:11:21 +00:00
Hieu Hoang	a27f16972b	correct active sentences vector size	2018-01-22 13:44:45 +00:00
Hieu Hoang	af64277672	batch stats	2018-01-21 22:36:05 +00:00
Hieu Hoang	1549edefa8	don't use shape for now but leave code in	2018-01-20 16:44:39 +00:00
Hieu Hoang	d88cbf7f82	shrink size of shape in Broadcast	2018-01-20 16:31:57 +00:00
Hieu Hoang	04fe617903	get ready to shrink size of shape in Broadcast	2018-01-20 16:26:17 +00:00
Hieu Hoang	434303d233	debug Broadcast - show max length and active batch max length	2018-01-20 16:06:44 +00:00
Hieu Hoang	7e78a6a155	timings	2018-01-19 19:03:17 +00:00
Hieu Hoang	f1033ac281	topup whenn half empty	2018-01-19 17:08:49 +00:00
Hieu Hoang	79de7a566e	check last cuda error after running every kernel. Program may not be compiled for the particular GPU or shared mem incorrectly set	2018-01-19 16:59:15 +00:00
Hieu Hoang	736e18ad5f	don't store hypo and vocab index in NthOutBatch	2018-01-19 13:36:11 +00:00
Hieu Hoang	8ca6783e2b	debug	2018-01-19 13:24:32 +00:00
Hieu Hoang	6a21025838	runs on gtx 1060	2018-01-19 12:11:29 +00:00
Hieu Hoang	c62c7619b8	debug	2018-01-18 20:34:18 +00:00
Hieu Hoang	70fd8dce60	debug	2018-01-18 19:55:24 +00:00
Hieu Hoang	1d6cc7cacd	correct function names	2018-01-18 19:40:46 +00:00
Hieu Hoang	2163b02237	delete cublasSetMathMode in handles.cu	2018-01-18 17:50:56 +00:00
Hieu Hoang	6cdb4b7511	Merge branch 'const-batch2' of https://github.com/hieuhoang/marian into const-batch2	2018-01-18 17:38:43 +00:00
Hieu Hoang	2f8751c20d	Merge branch 'const-batch2' of github.com:hieuhoang/marian into const-batch2	2018-01-18 17:38:26 +00:00
Hieu Hoang	d430350f2d	add --tensor-cores arg	2018-01-18 17:38:09 +00:00
Hieu Hoang	83b6de15c5	add tensorcore support	2018-01-16 16:03:06 +00:00
Hieu Hoang	4a725ed4ec	move syncthread out of if statement. Locks up on volta	2018-01-16 12:21:56 +00:00
Hieu Hoang	23b73264d1	formatting	2018-01-12 00:35:54 +00:00
Hieu Hoang	222de7dae0	Merge ../marian.hieu into const-batch2	2018-01-11 22:34:13 +00:00
Hieu Hoang	eb55a53944	no tensor core	2018-01-11 22:33:48 +00:00
Hieu Hoang	772df1fb0e	timings	2018-01-11 22:23:51 +00:00
Hieu Hoang	524d9ba968	swap sentenceLengths, not copy	2018-01-11 22:16:18 +00:00
Hieu Hoang	85c9b02f6c	use copyFrom	2018-01-11 22:07:04 +00:00
Hieu Hoang	72c2caa405	thread_local variables in TopupBatch	2018-01-11 21:59:03 +00:00
Hieu Hoang	28e71c94ff	create copyFrom	2018-01-11 21:42:22 +00:00
Hieu Hoang	bfd0a36758	cleanup	2018-01-11 21:23:40 +00:00
Hieu Hoang	1254853d38	don't copy if no rows specified for CopyMatrix3	2018-01-11 21:58:36 +01:00
Hieu Hoang	6414036d1a	don't copy if no rows specified	2018-01-11 21:36:49 +01:00
Hieu Hoang	712eb1b720	enable tensor core	2018-01-09 22:32:02 +00:00
Hieu Hoang	f9275f43ef	move .h -> .cu	2018-01-09 22:28:19 +00:00
Hieu Hoang	bbb396c303	minActive for different beam sizes	2018-01-08 23:19:38 +00:00
Hieu Hoang	db4533ae26	minActive for different beam sizes	2018-01-08 23:13:52 +00:00
Hieu Hoang	dd8beb59a5	minActive for different beam sizes	2018-01-08 23:12:43 +00:00
Hieu Hoang	c9b301772e	TIME_CMD() macro. Tweak SentencesToGet()	2018-01-08 22:50:49 +00:00
Hieu Hoang	2f807f7af1	tweak SentencesToGet()	2018-01-08 22:06:21 +00:00