f054ca281f
fixed Linux Makefile wrt deleted files
2016-01-14 09:06:40 +01:00
88cf193a50
Disable popups on Windows
2016-01-13 10:11:28 -08:00
24900d073f
made gcc happy
2015-12-28 20:03:50 -08:00
02d0405118
Source/Readers/LMSequenceReader/: also build SequenceWriter on Linux
2015-12-18 11:59:40 +00:00
ada359aecd
Fix for LibSVMBinaryReader to add prefetching, microbatches, and Linux support.
2015-12-16 12:13:38 -08:00
88db0cc804
Makefile: adapt to new location of generate_build_infop script
2015-12-16 14:24:49 +00:00
4faea2131e
Adapt Linux build
2015-12-16 08:05:32 +00:00
c918f3b6b8
started the TensorView class by adding header and CPP to math lib
2015-12-15 15:02:43 -08:00
5bd7852956
Fix kaldi Reader. 1) Using the new MBLayout interface. 2) Modify the configParameter to configRecordType to consistent with HTKMLFReader. 3) Clean the warning message. Next: refactor to make it consistent with HTKMLFReader.
2015-12-14 11:00:20 -05:00
72623a809f
cudnn: added OpenCV libs to Linux build.
2015-12-12 10:00:01 -08:00
3472e54524
cudnn: enabled build on Linux with cuDNN.
2015-12-12 09:59:09 -08:00
ec4df193d2
renamed commandArgUtil.h to Config.h and ConfigFile.cpp to Config.cpp;
...
bug fix in previous refactoring of sub-minibatching
2015-12-04 22:31:47 -08:00
83f68a3fb0
Step 2 of moving LSTMNode to EsotericNodes.h: deleted it from RecurrentNodes.h;
...
made gcc happy
2015-12-04 20:56:33 -08:00
1f66a70bd0
Added NDEBUG to Linux Release build.
2015-12-03 11:20:50 -08:00
bf071da9a8
add NVMLPATH to the linking target cntk for the libnvidia-ml.so
2015-11-26 07:34:37 -05:00
615308c219
Added call stack printing functionality for Linux
2015-11-25 17:20:56 -08:00
a5afd67610
Fix a bug in Scripts/genrate_build_info
2015-11-24 21:34:30 -08:00
cae798449f
Enable PrintBuildInfo for both Windows and Linux
2015-11-24 16:38:37 -08:00
c14149f7d6
simplified DeviceFromConfig() by removing PTasks-specific configuration options. Also added an overload for BrainScript;
...
added a dummy DataReader constructor from BrainScript (currently not implemented);
had to disable DelayLoadNofify() as that suddenly caused link errors (seems because DeviceFromConfig() is now called from a different place?);
further unified DoTrain() w.r.t. old CNTK config and BrainScript. Added a new BrainScriptBuilder (but the builders will go away next and get replaced by a lambda). DoTrain now accessible from BS (but not functional yet, too many pieces missing);
all builders now can be constructed from both old ConfigParameters and BrainScript IConfigRecord;
suddenly gcc got very picky, had to fix several places
2015-11-21 00:16:36 -08:00
e0d900f814
added more scaffolding (empty CPP files) for moving actions to CNTKActionsLib. Updated Linux Makefile
2015-11-17 23:22:33 -08:00
184485885b
Eliminated NetworkBuilderFromConfig.cpp. Instead, the few surviving lines--the lines registering ComputationNode as a scriptable object--are now in ComputationNode.cpp;
...
fixed Linux Makefile after change of GPUMatrixCUDAKernels.cu to GPUMatrixCUDAKernels.cuh. It used to explicitly compile this in the GCC build, although it was included
2015-11-17 22:37:43 -08:00
2ca3f555a8
BrainScript: ComputationNetwork can now construct itself from BS (the special MakeRuntimeObject() function is gone);
...
new source file ComputationNetworkScripting.cpp;
changed ComputationNetwork::SetDeviceId(). Now just takes the input and sets that as the device. Before it did somethinf funky with m_deviceId, which looks like a bug;
in LSTM test config, removed remnants of old NDL, now completely based on BS
2015-11-16 09:57:01 -08:00
1a423edcb9
added CNTKActionsLib to Linux include path;
...
moved DisableLegacyTruncationSettings() away from being the very first piece of code in CNTK.cpp, hiding it close to where it is used. Also fixed its spelling and made it 'static'
2015-11-12 13:23:54 -08:00
b1f0ad70ba
Fix two perf problems
...
1. On Linux, use ACML MP as default ACML library.
2. For the parallel end-to-end tests, limit concurrency based on
available hardware threads.
2015-11-05 22:15:51 +01:00
1b5fa2dfde
Applying workaround for race condition in OpenMPI on non-0 exit code from CNTK
2015-10-30 16:16:09 -07:00
317b9466cd
moved memshare functions from header to ComputationNetworkEvaluation.cpp;
...
created new ComputationNetworkEditing.cpp
2015-10-30 10:35:12 -07:00
aa0fb1db80
moved sets of functions from ComputationNetwork.cpp to ComputationNetworkEvaluation.cpp and ComputationNetworkAnalysis.cpp
2015-10-30 10:05:12 -07:00
cabada4f8d
Fix build for NoGPU.
2015-10-28 20:59:52 -04:00
3343afb092
Remove debug info generation option for CUDA compilation in debug flavor builds and also enable fast-math optimizations . These changes have been done to eliminate differences in GPU results for the E2E tests between debug and release flavors. Setting environment variable CNTK_CUDA_DEVICE_DEBUGINFO=1 will enable debug info generation. The baselines for all E2E tests have also been updated in accordance with this change
2015-10-24 14:34:01 -07:00
5c36007848
Rename SequenceTraining project to CNTKSequenceTrainingLib to be consistent with other libs
2015-10-18 12:40:43 -07:00
e065d46a9b
ImageReader fixes for Linux build.
2015-10-14 12:30:11 -07:00
83026855e5
Add CUB path.
2015-10-14 12:29:51 -07:00
0155e9e195
Fixed Linux build
2015-10-09 23:54:37 -07:00
b415924234
Merge remote-tracking branch 'remotes/origin/master' into guoguo/linuxBuildFix
2015-10-07 18:19:11 -04:00
32eedca2b6
Refactor SequenceTraining to support CPUONLY.
2015-10-07 10:35:31 -07:00
bc7b7ae63f
more cleanup on when Resize() ops are done;
...
renamed SimpleEvaluator::PreCompute() to EvaluateBatchModeNodes() because the original name was misleading;
Linux Makefile now defines _DEBUG in debug builds
2015-10-06 14:02:45 -07:00
8ae7966876
fixing linux build for master branch; removed duplicated files in KaldiReader and Kaldi2Reader
2015-09-30 18:02:05 -04:00
47cb3ce1bd
Port the Sequence trainign code to Linux
2015-09-25 17:22:58 -07:00
681e060880
Merge branch 'master' of https://git01.codeplex.com/forks/srehpycs/mitsls into srehpycs/mitsls
2015-09-24 10:50:00 -07:00
577556cfdc
Add env vars to control CUDA codegen.
2015-09-17 16:57:26 -07:00
dede055e1c
Change CUDA settings for Debug builds.
2015-09-17 16:37:00 -07:00
409ca3b456
Eradicate all possible warnings from g++
...
Replace hash_map with unordered_map
Fix size_t formatting (%zd)
Add some missing sprintf_s args
Some methods defined in base class need the base mentioned in templates
Don't use classname::methodname in class definitions
2015-09-17 16:00:48 -04:00
cf29bf0f38
Merge remote-tracking branch 'cntk/master' into merge
2015-09-17 09:32:17 -04:00
62558e3079
Moved the parallel trainign guard when writing model/checkpoint files to the actual save functions instead of guarding at the call sites
2015-09-15 21:04:09 -07:00
9aecb5649d
made gcc happy again (mostly missing headers or wrong declaration orders);
...
Makefile adapted to new paths, but not yet building Network and SGD as separate libs
2015-09-06 09:20:28 -07:00
7f4e085958
CNTKEval now builds (missed the new .cpp files)
2015-09-03 19:15:00 -07:00
9a6eb2ac00
moved most <ElemType>-dependent functions out of ComputationNetwork into a ComputationNetworkBuilder class (added two files);
...
fixed the slew of dependencies, mostly SimpleNetworkBuilder.cpp;
ComputationNetwork no longer has to include all headers for all vairants of ComputationNodes;
a few ComputationNode derivates missed #includes
2015-09-03 18:41:59 -07:00
38791df522
Merge branch 'master' into fseide/bs
2015-09-03 15:05:46 -07:00
785429e854
Merge branch 'master' into fseide/bs
...
Conflicts:
Math/Math/NoGPU.cpp
2015-09-03 13:50:03 -07:00
fb0ed55e6e
NVCC warnings as errors
2015-09-03 13:30:30 -07:00
Philipp Kranen
Marko Radmilac
Frank Seide
Mark Hillebrand
thhoens
yzhang87
Alexey Kamenev
Qiwei Ye
Amit
Yongqiang Wang
Vladimir Ivanov
Alexey
Guoguo Chen
Scott Cyphers
U-FAREAST\fseide