I have success installed theano 0.82 and cuda 7.5，and now theano can compute with gpu.
But now there is a problem, every time when I launch a new python process and import theano, nvcc would always generate libraries , even I did not change any theanorc config. It takes more than 10 seconds each time, when I run a small program, it waste too much more than the time saved by gpu. Python 2.7.12 |Anaconda 4.1.1 (64-bit)| (default, Jun 29 2016, 11:07:13) [MSC v1500 64 bit (AMD64)] on win32 >>> import theano DEBUG: nvcc STDOUT mod.cu Creating libraris: C:/Users/Administrator/AppData/Local/Theano/compiledir_Windows-7-6 .1.7601-SP1-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.12-64/ tmpzymq1 z/265abc51f7c376c224983485238ff1a5.lib and object C:/Users/Administrator/AppData/Loc al/Theano/compiledir_Windows-7-6.1.7601-SP1-Intel64_Family_6_Model_60_Stepping_3 _GenuineIntel-2.7.12-64/tmpzymq1z/265abc51f7c376c224983485238ff1a5.exp Using gpu device 0: GeForce GTX 780 Ti (CNMeM is enabled with initial size: 60.0 % of memory, cuDNN 5005) Here is another import: >>> import theano DEBUG: nvcc STDOUT mod.cu Creating libraris: C:/Users/Administrator/AppData/Local/Theano/compiledir_Windows-7-6 .1.7601-SP1-Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-2.7.12-64/ tmpzw9yb b/265abc51f7c376c224983485238ff1a5.lib and object C:/Users/Administrator/AppData/Loc al/Theano/compiledir_Windows-7-6.1.7601-SP1-Intel64_Family_6_Model_60_Stepping_3 _GenuineIntel-2.7.12-64/tmpzw9ybb/265abc51f7c376c224983485238ff1a5.exp The generated lib files and exp files have same filenames Is this step necessary? I think the generated libraries should be reuse when nothing changed, How can I avoid this ? some nvcc flags config in theanorc? Or this is a bug of nvcc or theano？ Thanks!! -- --- 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.
