merge from 8.x to master

.gitignore

@@ -3,6 +3,7 @@
 *.lo
 *.o
 *.obj
+foo.dat
 
 # Precompiled Headers
 *.gch

CMakeLists.txt

@@ -13,12 +13,25 @@ if (NOT CMAKE_BUILD_TYPE)
     set(CMAKE_BUILD_TYPE "RelWithDebInfo" CACHE STRING "Choose the type of build, options are: Debug Release RelWithDebInfo MinSizeRel." FORCE)
 endif()
 
-OPTION(LIBJPEG_AVAILABLE "Allows native loading of jpegs, via manifest file." ON)
-OPTION(DEV_RUN_COG "DeepCL maintainers only, otherwise set to 'OFF'." OFF)
-OPTION(PROTOTYPING "Only for devs." OFF)
-    mark_as_advanced( FORCE PROTOTYPING )
-OPTION(LIBPNGPP_AVAILABLE "Some toys/samples only." OFF)
-    mark_as_advanced(FORCE LIBPNGPP_AVAILABLE)
+#OPTION(BUILD_PYTHON_WRAPPERS "Build Python wrappers.  Needs Python." ON)
+OPTION(BUILD_JPEG_SUPPORT "Allows native loading of jpegs, via manifest file." ON)
+OPTION(BUILD_INTERNAL_LUA "If using from Lua, set to 'OFF'" ON)
+OPTION(MAINTAINER_OPTIONS "Show maintainer options" OFF)
+
+if(MAINTAINER_OPTIONS)
+    OPTION(BUILD_PYTHON_WRAPPERS "Build python wrappers.  Maintainers only." OFF.)
+    OPTION(DEV_RUN_COG "DeepCL maintainers only, otherwise set to 'OFF'." OFF)
+    OPTION(PROTOTYPING "Only for devs." OFF)
+#    mark_as_advanced( FORCE PROTOTYPING )
+    OPTION(LIBPNGPP_AVAILABLE "Some toys/samples only." OFF)
+#    mark_as_advanced(FORCE LIBPNGPP_AVAILABLE)
+else()
+    unset(DEV_RUN_COG CACHE)
+    unset(DEV_RUN_CYTHON CACHE)
+    unset(BUILD_PYTHON_WRAPPERS CACHE)
+    unset(LIBPNGPP_AVAILABLE CACHE)
+    unset(PROTOTYPING CACHE)
+endif()
 
 IF(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
   SET(CMAKE_INSTALL_PREFIX
@@ -27,9 +40,10 @@ IF(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
 ENDIF(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
 
 # remove old flags from the cache:
+unset(LIBJPEG_AVAILABLE CACHE)
+unset(PROVIDE_LUA_ENGINE CACHE)
 unset(BUILD_PYSWIG_WRAPPERS CACHE)
 unset(BUILD_LUA_WRAPPERS CACHE)
-unset(BUILD_PYTHON_WRAPPERS CACHE)
 unset(RUN_COG CACHE)
 unset(RUN_SWIG CACHE)
 unset(DEV_RUN_SWIG CACHE)
@@ -42,7 +56,7 @@ unset(LUA_INCLUDE_DIR CACHE)
 
 # SET(CMAKE_INSTALL_PREFIX "${CMAKE_CURRENT_SOURCE_DIR}/dist" CACHE STRING "Installation directory." FORCE)
 
-SET(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
+SET(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake")
 
 INCLUDE("${CMAKE_MODULE_PATH}/build_EasyCL.cmake")
 INCLUDE_DIRECTORIES(${EASYCL_INCLUDE_DIRS})
@@ -57,15 +71,14 @@ include_directories( src )
 include_directories( qlearning )
 
 if( ON_LINUX )
-#    target_link_libraries(EasyCL dl)
     SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++0x -Wall")
 endif()
 
 if( ON_WINDOWS)
     link_libraries(winmm) # needed for timeGetTime
 endif()
 
-set(dirs activate batch clmath conv dropout fc forcebackprop input layer loaders
+set(dirs clblas activate batch clmath conv dropout fc forcebackprop input layer loaders
    loss net netdef normalize patches pooling trainers util weights qlearning
    )
 foreach(dir ${dirs})
@@ -75,6 +88,7 @@ foreach(dir ${dirs})
         set( deepcl_sources ${deepcl_sources} src/${dir}/${source})
     endforeach()
 endforeach()
+set(deepcl_sources ${deepcl_sources} src/DeepCL.cpp src/CppRuntimeBoundary.cpp)
 #message("deepcl_sources ${deepcl_sources}")
 
 #find_package(Lua51)
@@ -141,23 +155,25 @@ if( LIBPNGPP_AVAILABLE)
     target_link_libraries(deepcl_unittests ${PNG_LIBRARY})
 endif(LIBPNGPP_AVAILABLE)
 
-link_libraries(DeepCL)
+#link_libraries(DeepCL)
 
 if(DEV_RUN_COG)
     foreach( dir ${dirs} main)
         set(cog_dirs ${cog_dirs} ${CMAKE_CURRENT_SOURCE_DIR}/src/${dir}/*.cpp ${CMAKE_CURRENT_SOURCE_DIR}/src/${dir}/*.h)
     endforeach()
     add_custom_target(
         cog
-        python ${CMAKE_CURRENT_SOURCE_DIR}/thirdparty/cogapp/cog.py -q -I ${CMAKE_CURRENT_SOURCE_DIR}/cog-batteries -r ${cog_dirs} ${CMAKE_CURRENT_SOURCE_DIR}/test/*.cpp ${CMAKE_CURRENT_SOURCE_DIR}/test/*.h ${CMAKE_CURRENT_SOURCE_DIR}/prototyping/*.cpp
+        python ${CMAKE_CURRENT_SOURCE_DIR}/thirdparty/cogapp/cog.py -q -I ${CMAKE_CURRENT_SOURCE_DIR}/cog-batteries -r ${cog_dirs} ${CMAKE_CURRENT_SOURCE_DIR}/test/*.cpp ${CMAKE_CURRENT_SOURCE_DIR}/test/*.h ${CMAKE_CURRENT_SOURCE_DIR}/prototyping/*.cpp ${CMAKE_CURRENT_SOURCE_DIR}/src/*.h
         WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
     )
     add_dependencies( DeepCL cog )
 endif(DEV_RUN_COG)
 
 #ADD_DEPENDENCIES(DeepCL clBLAS)
 
-set( UNITTEST_SOURCES test/testupdateweights.cpp test/testforward.cpp test/testfilehelper.cpp
+set( UNITTEST_SOURCES test/testClBlas.cpp
+ test/testDeepCL.cpp
+ test/testupdateweights.cpp test/testforward.cpp test/testfilehelper.cpp
  test/testsimpleconvolvenet.cpp test/testlogicaloperators.cpp 
  test/testbackward.cpp test/testsinglebatch.cpp 
  test/testpoolingforward.cpp test/testpoolingbackward.cpp test/testNorbLoader.cpp
@@ -173,6 +189,7 @@ set( UNITTEST_SOURCES test/testupdateweights.cpp test/testforward.cpp test/testf
 if(LIBJPEG_AVAILABLE)
     set(UNITTEST_SOURCES ${UNITTEST_SOURCES} test/testjpeghelper.cpp)
     add_executable( mnist-to-jpegs test/mnist-to-jpegs.cpp src/util/stringhelper.cpp src/loaders/MnistLoader.cpp )
+    target_link_libraries(mnist-to-jpegs DeepCL)
 endif(LIBJPEG_AVAILABLE)
 
 add_executable( deepcl_train src/main/train.cpp src/util/stringhelper.cpp )
@@ -183,13 +200,21 @@ add_executable( prepare-norb test/prepare-norb.cpp src/util/stringhelper.cpp )
 add_executable( mnist-to-floats test/mnist-to-floats.cpp src/util/stringhelper.cpp )
 add_executable( mnist-to-pipe test/mnist-to-pipe.cpp src/util/stringhelper.cpp )
 
+foreach(exe deepcl_train deepcl_predict cifar-to-mat prepare-norb mnist-to-floats mnist-to-pipe)
+    target_link_libraries(${exe} DeepCL)
+endforeach()
+
 #target_link_libraries( cifar-to-mat ${LUA_LIBRARIES} )
 
 if( LIBPNGPP_AVAILABLE)
     add_definitions(-DPNG_AVAILABLE)
     add_executable( testpatchextractor test/testPatchExtractor.cpp src/util/stringhelper.cpp src/patches/PatchExtractor.cpp )
     add_executable( testtranslator test/testTranslator.cpp src/util/stringhelper.cpp src/patches/Translator.cpp )
     add_executable( testgenericloader test/testGenericLoader.cpp src/util/stringhelper.cpp )
+
+    foreach(exe testpatchextractor testtranslator testgenericloader)
+        target_link_libraries(${exe} DeepCL)
+    endforeach()
 endif(LIBPNGPP_AVAILABLE)
 
 if( ON_LINUX )
@@ -215,9 +240,17 @@ target_link_libraries( deepcl_unittests DeepCL )
 add_executable( testgtestsupp test/testgtestsupp.cpp thirdparty/gtest/gtest_main.cc )
 target_link_libraries( testgtestsupp deepcl_gtest )
 
+if(BUILD_PYTHON_WRAPPERS)
+    add_subdirectory(python)
+#    add_dependencies(PyDeepCL DeepCL)
+#    target_link_libraries(PyDeepCL DeepCL)
+endif()
+
 INSTALL( DIRECTORY src/ DESTINATION include/deepcl FILES_MATCHING PATTERN *.h )
+INSTALL(PROGRAMS src/activate.sh DESTINATION bin)
+INSTALL(PROGRAMS src/activate.bat DESTINATION bin)
 #INSTALL( DIRECTORY EasyCL/ DESTINATION include/easycl FILES_MATCHING PATTERN *.h )
-INSTALL( TARGETS DeepCL deepcl_train deepcl_predict deepcl_unittests
+INSTALL( TARGETS DeepCL deepcl_train deepcl_predict deepcl_unittests deepcl_gtest
     RUNTIME DESTINATION bin
     ARCHIVE DESTINATION lib
     LIBRARY DESTINATION lib )

README.md

@@ -76,34 +76,57 @@ Multicolumn net also possible, as in [McDnn](http://arxiv.org/pdf/1202.2745.pdf)
 - obtained 99.5% test accuracy on MNIST, using `netdef=rt2-8c5z-relu-mp2-16c5z-relu-mp3-150n-tanh-10n numepochs=20 multinet=6 learningrate=0.002`
    - epoch time 99.8 seconds, using an Amazon GPU instance, ie half an NVidia GRID K520 GPU (since we are learning 6 nets in parallel, so 16.6seconds per epoch per net)
 
-# Releases
+# Installation
 
-* Stable: here
-* Unstable/dev: please see [8.x](https://github.com/hughperkins/DeepCL/tree/8.x)
+## Native library installation
 
-Unstable/dev is working on adding im2col, which promises faster convolution for larger layers, ie layers 1 and 2 in [Soumith's benchmarks](https://github.com/soumith/convnet-benchmarks)
+This section installs the native libraries, and the command-line tools.  You always need to do this part, even if you will use the Python wrappers.
 
-# To install
+### Windows
 
-## Python
+#### Pre-requisites:
 
-* For python, please use [Python API](python/README.md), or use [pip](https://pypi.python.org/pypi/DeepCL)
+* OpenCL-enabled GPU or APU, along with appropriate OpenCL driver installed
+* Tested using Windows 7
 
-## Commandline tools, and c++ libraries
+#### Procedure:
 
-### Windows
+* Download latest binary zip file from http://deepcl.hughperkins.com/Downloads/
+* unzip it, which creates the `dist` folder
+* To use it:
+  * open a cmd
+  * run `call dist\bin\activate.bat` (adjusting the path appropriately for wherever you downloaded deepcl binaries to)
+  * now, eg try `deepcl_unittests`
+
+Note that you need to "activate" the installation each time you open a new cmd prompt (or you could add appropriate environment variables permanently, using Control Panel | System | Advanced System Settings | Environment Variables)
+
+### Linux
+
+#### Pre-requisites:
+
+* OpenCL-enabled GPU or APU, along with appropriate OpenCL driver installed (can check by running `clinfo`, which should show your desired GPU device)
+* Tested using Ubuntu 14.04 32-bit/64-bit
+
+#### Procedure:
 
-Pre-built binaries are available for Windows.  In order to use them you need:
-* An OpenCL driver for your GPU
-* A recent release with Windows binaries is [v5.5.0](https://github.com/hughperkins/DeepCL/releases/tag/v5.5.0) 
+* Download latest tar file from http://deepcl.hughperkins.com/Downloads/
+* untar it, which creates the `dist` sub-folder
+* in a bash prompt, run `source dist\bin\activate.sh` (adjust the path appropriate for wherever you untarred the binaries tar file to)
+* test by doing, from the same bash prompt, eg `deepcl_unittests`
 
-### linux
+Note that you need to "activate" the installation each time you open a new bash prompt (or you can call activate.sh from your `.bashrc` file)
 
-Pre-build binaries are available for linux.  In order to use them you need:
-* An OpenCL driver for your GPU
-* A recent release with linux binaries is [v5.5.0](https://github.com/hughperkins/DeepCL/releases/tag/v5.5.0) 
+## Python wrappers
 
-If the binaries dont work on your distribution, please [build from source](doc/Build.md)
+* make sure you already installed the native library, and "activate"d it, by doing `call dist\bin\activate.bat`, or `source dist/bin/activate.sh`
+* run `pip install --pre DeepCL`
+* test by doing `python -c "import PyDeepCL; cl = PyDeepCL.DeepCL()"`
+
+## To build from source
+
+Building from source is only needed if installing from binaries doesn't work for your configuration, or if you want to modify DeepCL.
+
+See [Build.md](doc/Build.md)
 
 ## What if it doesn't run?
 
@@ -137,11 +160,8 @@ Related projects
 ================
 
 * [kgsgo-dataset-preprocessor](https://github.com/hughperkins/kgsgo-dataset-preprocessor) Dataset based on kgsgo games; 33 million data points
-
-Credits
-=======
-
-* Tambet Matilsen has provided excellent suggestions and feedback on which functionalities to prioritize, and on how to make the website somewhat presentable
+* [cltorch](https://github.com/hughperkins/cltorch)
+* [clnn](https://github.com/hughperkins/clnn)
 
 License
 =======
@@ -151,12 +171,66 @@ License
 Recent changes
 ==============
 
+* Aug 28th:
+  * installation of 8.x from binaries on Windows works now, by doing, eg on 32-bit Windows 7, and assuming you already activated an appropriate python environment (assumes 7-zip is installed, in default location, otherwise do the unzip by hand):
+```
+powershell Set-ExecutionPolicy unrestricted
+rem following command is like `wget` in linux:
+powershell.exe -Command (new-object System.Net.WebClient).DownloadFile('http://deepcl.hughperkins.com/Downloads/deepcl-win32-v8.0.0rc8.zip', 'deepcl-win32-v8.0.0rc8.zip')
+rem following command is like `tar -xf` in linux:
+"c:\program files\7-Zip\7z.exe" x deepcl-win32-v8.0.0rc8.zip
+call dist\bin\activate.bat
+pip install --pre DeepCL
+python -c "import PyDeepCL; cl = PyDeepCL.DeepCL()"
+```
+(last line is just to check works ok)
+  * merged 8.x branch to master, will release first version of 8.x shortly
+* Aug 26th: installation of 8.x from binaries on linux works now, by doing, eg on 64-bit Ubuntu 14.04:
+```
+mkdir 8.0.0rc4
+cd 8.0.0rc4
+wget http://deepcl.hughperkins.com/Downloads/deepcl-linux64-v8.0.0rc4.tar.bz2
+tar -xf deepcl-linux64-v8.0.0rc4.tar.bz2
+virtualenv env
+source env/bin/activate
+source dist/bin/activate.sh
+pip install --pre DeepCL
+python -c "import PyDeepCL; cl = PyDeepCL.DeepCL()"
+```
+(last line is just to check works ok)
+* Aug 21st-24th:
+  * 8.x finally builds again on all supported configurations!
+    * ubuntu 14.04 32-bit Python 2.7
+    * ubuntu 14.04 32-bit Python 3.4
+    * ubuntu 14.04 64-bit Python 2.7
+    * ubuntu 14.04 64-bit Python 3.4
+    * visual studio 2010 32-bit python 2.7
+    * visual studio 2010 32-bit python 3.4
+    * visual studio 2010 64-bit python 2.7
+    * visual studio 2010 64-bit python 3.4
+* Aug 19th-20th:
+  * Python wrappers now built using a very thin setup.py layer, on top of the standard native DeepCL build
+* Aug 18th:
+  * added BackwardIm2Col layer, which uses im2col for backward propagation
+  * added BackpropWeightsIm2Col layer, which uses im2col for weight update
+  * added BackwardAuto layer, which automatically selects fastest Backward layer
+  * added BackpropWeightsAuto layer, which automatically selects faster weight update layer
+  * under the covers:
+    * created ClBlasHelper, to handle Gemm and Gemv
+    * factorized im2col into Im2Col class
+* week up to Aug 17th:
+  * added forward and backward im2col layer
+  * forward im2col automatically used during forward propagation, where appropriate
+  * backwards has yet to be integrated
+  * under the covers:
+    * added clBLAS
+    * migrated the Python build process to use cmake, rather than setup.py (whether this turns out to be good or bad is a bit up in the air for now)
 * June 22nd:
   * removed lua wrappers
   * if you want to use lua with OpenCL, please consider using [cltorch](http://github.com/hughperkins/cltorch) and [clnn](http://github.com/hughperkins/clnn)
 
 To get in contact
 =================
 
-There is a mailing list at http://lists.hughperkins.com/listinfo.cgi/deepcl-hughperkins.com for discussions, ideas, or just to say 'hi'.  You can also just create issues, in github, in the top right of this page.
+Just create an issues, in github, in the top right of this page.  Don't worry about whether you think the issue sounds silly or anything.  The more feedback the better!