Update:

I found someone who helped me get things installed so I could at least see 
things in PyCharm and then after a lot of Google searching and playing 
around I finally have a setup that has lasagne/theano working in PyCharm.  
I'm on a laptop without a gpu on Windows 7.  Here's what worked for me 
(times are check_blas.py):

1) Installed Anaconda.  It now includes a BLAS called mkl (35s) that is 
faster than OpenBLAS (55s).  I have 3.5 but if using nolearn you should use 
3.4 since nolearn won't work with 3.5 per the pip error.

2) Created a virtual environment for the network.

3) From the theano install tutorial ( 
http://deeplearning.net/software/theano/install_windows.html#install-windows ) 
installed Visual Studio 10 Express from the all-in-one cd link.  Had to 
download a cd reader as directed.  Also installed TDM-GCC-64 from that 
site. 


4) From the site that Jan put together ( 
https://github.com/Lasagne/Lasagne/wiki/From-Zero-to-Lasagne-on-Windows-7-(64-bit)
 
), installed CUDA, Windows SDK

5) Installed Mingw64 (This may or may not have been automatically installed 
with Anaconda but I have them).  I also installed the Visual C++ for Python 
listed in the theano tutorial

6) Updated the nvcc.profile file in the C:\Program Files\NVIDIA GPU 
Computing Toolkit\CUDA\v5.5\bin directory to read:


TOP              = $(_HERE_)/..
NVVMIR_LIBRARY_DIR = $(TOP)/nvvm/libdevice
PATH            += $(TOP)/open64/bin;$(TOP)/nvvm/bin;$(_HERE_);$(TOP)/lib;
INCLUDES        +=  "-I$(TOP)/include" $(_SPACE_)
INCLUDES += "-I$(TOP)/include" "-I$(TOP)/include/cudart" "-IC:/Program 
Files (x86)/Microsoft Visual Studio 10.0/VC/include" $(_SPACE_)
LIBRARIES        =+ $(_SPACE_) "/LIBPATH:$(TOP)/lib/$(_WIN_PLATFORM_)"
CUDAFE_FLAGS    +=
OPENCC_FLAGS    +=
PTXAS_FLAGS     +=

7) Updated .theanorc in the C:\Users\Anon directory  .Note to not use the 
force_device = true if you have a gpu, without that it prints out a long 
function and throws an error starting with "ERROR (theano.sandbox.cuda): 
Failed to compile cuda_ndarray.cu: ('nvcc return status',  ".  Without the 
blas section the speed slows to 70s:

[global]
force_device = True
device = cpu
floatx = float32
warn_float64 = warn

[blas]
ldflags = -LC:\Anaconda3\envs\corrnet\Library\bin -lmkl_rt

[gcc]
cxxflags = -LC:\Anaconda3\envs\corrnet\Library\include 
-LC:\Anaconda3\envs\corrnet\Library\lib -lm

[nvcc]
flags=--use-local-env --cl-version=2010 -D_FORCE_INLINES 
fastmath = True


8) I couldn't get theano to find the Mingw64 dll files and their 
dependencies ('mkl_core.dll', 'mkl_intel_lp64.dll', 'mkl_intel_thread.dll') 
so just copied the whole 
C:\Anaconda3\envs\ainn\MinGW\x86_64-w64-mingw32\include 
directory into  C:\Anaconda3\envs\ainn\include including copying the 
subdirectories intact.  There's probably a more elegant way but this works 
so don't care at this point.

9) My path and system variables (found by right-click computer, properties, 
advanced system settings, environmental variables) probably has too much in 
it but has:  

HOME   %USERPROFILE%
PYTHONPATH   C:\Anaconda3\;C:\Anaconda3\Lib;

The CUDA paths installed by CUDA

PATH   C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5;
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5\bin;
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5\libnvvp;
C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin;
C:\Program Files\Microsoft Visual Studio 10.0\VC\bin\amd64;
C:\Program Files\Microsoft Visual Studio 10.0\Common7\IDE;
c:\Program Files (x86)\Microsoft SQL Server\100\Tools\Binn;
c:\Program Files\Microsoft SQL Server\100\Tools\Binn;
c:\Program Files\Microsoft SQL Server\100\DTS\Binn;
C:\Program Files (x86)\QuickTime\QTSystem\;
C:\TDM-GCC-64\bin;
C:\Program Files (x86)\Microsoft SDKs\Windows\v7.0A\Lib;
C:\Anaconda3\;C:\Anaconda3\bin;
C:\Anaconda3\lib;
C:\Anaconda3\libs;
C:\Anaconda3\scripts;
C:\Anaconda3\tools;
C:\Users\Anon\AppData\Local\Programs\Common\Microsoft\Visual C++ for 
Python\9.0\VC\bin;
C:\Users\Anon\AppData\Local\Programs\Common\Microsoft\Visual C++ for 
Python\9.0\VC\lib;
C:\Users\Anon\AppData\Local\Programs\Common\Microsoft\Visual C++ for 
Python\9.0\WinSDK\bin;
C:\Users\Anon\AppData\Local\Programs\Common\Microsoft\Visual C++ for 
Python\9.0\WinSDK\lib;
C:\Users\Anon\AppData\Local\Programs\Common\Microsoft\Visual C++ for 
Python\9.0;
C:\Program Files (x86)\Microsoft Visual Studio 10.0\Common7\IDE

10) In PyCharm created a new project that used the new Anaconda environment 
and made sure all relevant files were in the same directory.

Now the MNIST.py and check_blas.py files run fine and relatively fast for 
not having a GPU.

Anyways hopefully this helps someone else.  It was a nightmare. I haven't 
tried debugging yet but at least it's working in PyCharm now.  I'll update 
once I have something to debug.

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to