a06c88cf6a
added TempDeviceBuffer to CommonMatrix.h
2016-11-19 21:53:24 +01:00
913dcd31d4
fixes for dense times sparse
2016-11-19 21:53:24 +01:00
b555be9303
Incorporated CR feedback
2016-11-17 17:14:34 -08:00
4e88307aba
added a bunch of ops to load legacy models
2016-11-16 16:34:27 -08:00
88065644f7
Merge remote-tracking branch 'origin/master' into mahilleb/pr/874
2016-11-09 14:01:33 +00:00
b8d668ecf9
Source/Math/NcclComm.cpp: tune logging
2016-11-09 14:00:53 +00:00
814381b9a8
1. Fixed some known merging problems
...
2. Fixed a bug of EndForwardProp in BatchNormalization
2016-11-09 17:26:00 +08:00
4670158510
Added NCCL support to SimpleDistGradAggregator
2016-11-08 18:15:01 -08:00
01be41a2b3
Merge branch 'master' into v-yuxgu/memory-optv2
2016-11-08 16:14:12 +08:00
f5af601244
Supple comments and fix several bugs of memory optimization
2016-11-08 15:41:08 +08:00
9c02129e2b
First version of all reduce aggregator
2016-10-24 19:25:36 +02:00
e6036b2735
CNTK v2 library: Fix a bug in convolution inference and other misc bug fixes
2016-10-20 15:37:15 -07:00
f3289b0f84
First version of distributed aggregator for v2
2016-10-19 13:49:10 +02:00
feb3b127a7
minor change in definition of BatchNormEngine to avoid warnings during compilation with cuda8 compiler
2016-10-18 18:46:15 +02:00
c28ac3327e
.vcxproj: remove <BrowseInformation/> and <Bscmake/>
...
This is a partial revert of 18fa8097bf
2016-10-18 10:29:54 +02:00
d7e02247a8
Add Release_NoOpt flavor for python debugging
2016-10-14 14:21:23 -07:00
a183815dd1
enabling cross-compile of the cntk sources for arm64 devices.
...
This commit replaces implementations of functions based on sse with a non-sse version
and updates the makefile such that SSE_FLAGS can be overriden when compiling for non-Intel architectures.
2016-10-14 10:09:40 +02:00
279b25680c
Integrate wilrich/alpha4 into master
2016-10-13 16:44:40 -07:00
551682277f
Integrate fmegen/aarch64/new-gcc-5-4-0 into master
2016-10-13 10:27:52 -07:00
83bd64bcdd
Merge remote-tracking branch 'origin/master' into wilrich/alpha4
...
Conflicts:
bindings/python/cntk/utils/__init__.py
2016-10-13 12:56:03 +01:00
18fa8097bf
merge
2016-10-12 13:25:17 +02:00
a6bf86e6db
merged from master
2016-10-12 10:06:57 +02:00
12eeed3458
Fixing the GPUSparse to GPU matrix conversion
2016-10-11 14:26:30 +02:00
c5f25404ea
fix build errors due to new gcc compiler (5.x from ubuntu 16.4.1)
2016-10-07 22:00:02 +02:00
142af4bd7e
merging from master
2016-10-06 16:05:28 +02:00
901896b5ec
Do not invoke CachedDeviceProps() on Math.dll load
2016-10-06 11:38:00 +02:00
e0b62abf4a
Revert "Add NoOpt build flavor"
...
This reverts commit d8a14c9e2a
2016-10-05 14:07:01 -07:00
d8a14c9e2a
Add NoOpt build flavor
2016-10-05 13:53:14 -07:00
42d222aefc
Merge remote-tracking branch 'origin/master' into thilow/sampledSoftmax
...
# Conflicts:
# bindings/python/cntk/swig/swig.sh
2016-10-05 15:43:22 +02:00
e5429ae44d
Addressed BS CR comments
2016-10-04 14:55:10 +02:00
35bd0efc12
removed image layout from ROIPooling, changed float and double to ElemType
2016-10-04 14:55:05 +02:00
0f140ec735
Addressed CR comments round 2
2016-10-04 14:55:03 +02:00
e4bcf0f1b5
address review comments, round 1
2016-10-04 14:55:00 +02:00
6643c2b480
hunter: fast-rcnn
2016-10-04 14:54:59 +02:00
e446a8ebdc
merging from master
2016-10-04 09:50:48 +02:00
511ee243f0
Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/cntkv2Library
2016-10-02 22:12:12 -07:00
71f742ec61
CNTK v2 library: Few bug fixes, minor API changes and added scripts for ease of building/testing bindings from sources on Windows
2016-10-02 20:45:07 -07:00
284eaf189a
bug fix.
2016-10-02 11:29:25 -07:00
de1ddb18ae
Revert unpooling implementation to original version to be consistent with Caffe.
2016-10-02 11:29:25 -07:00
0094401323
Change max pooling to behave similarly with NVidia.
2016-10-02 11:29:25 -07:00
53fe8a36e7
Small fix for code review comments.
2016-10-02 11:27:36 -07:00
c2831aa02c
Bug fix for racing conditions in pooling, and handle multiple max values in max pooling and unpooling.
2016-10-02 11:26:26 -07:00
839810c0af
bug fix.
2016-10-02 11:23:27 -07:00
13ccc7f12c
Revert unpooling implementation to original version to be consistent with Caffe.
2016-10-02 11:23:27 -07:00
fd0cf7bb72
Change max pooling to behave similarly with NVidia.
2016-10-02 11:23:27 -07:00
9af240c066
Small fix for code review comments.
2016-10-02 11:23:27 -07:00
2db195e3f1
Bug fix for racing conditions in pooling, and handle multiple max values in max pooling and unpooling.
2016-10-02 11:23:27 -07:00
dcedf54c4d
merging from master
2016-09-27 18:04:46 +02:00
5ab3e0b012
undoing the move of valueMatrix.SwitchToMatrixType to validate
2016-09-27 14:54:01 +02:00
2e4a773398
Fixing empty lines
2016-09-27 13:36:31 +02:00
Thilo Will
Nikos Karampatziakis
Mark Hillebrand
yuxiao.guo
Nathan Luehr
Eldar Akchurin
Amit Agarwal
Wolfgang Manousek
KeDengMS
Friedel van Megen
Project Philly
Willi Richert
Alexey Reznichenko
Philipp Kranen
Hunter Lang
Cha Zhang