You have "optimizer = None" in your .theanorc, which prevents all graph optimizations from being applied, including the ones that replace the CPU computations by the GPU ones.
Theano is behaving normally. If you want a faster compilation but still using the GPU, use "optimizer = fast_compile". On Mon, Sep 26, 2016, ZetaOri wrote: > Hello. > > I am having an issue using the GPU with Theano. It says it runs on gpu, but > still performs computation on cpu. > > CentOS 6.7 > python 2.7.10 > cuda 7.5 > gcc 4.4.7. > > I have, as far as I know, configured everything correctly to run theano > with device=gpu, but it still runs on CPU. > > Running the theano gpu test > <http://deeplearning.net/software/theano/tutorial/using_gpu.html#testing-theano-with-gpu> > > outputs the following: > python tgputest.py > *Using gpu device 0: GeForce GTX TITAN X (CNMeM is disabled, cuDNN 4007)* > [HostFromGpu(<CudaNdarrayType(float32, vector)>), > Elemwise{exp,no_inplace}(HostFromGpu.0)] > Looping 1000 times took 4.698779 seconds > Result is [ 1.23178029 1.61879337 1.52278066 ..., 2.20771813 2.29967761 > 1.62323284] > *Used the cpu* > > theanorc is configured properly, but testing with THEANO_FLAGS shows exact > same output. > > What is frustrating is theano used to work on the gpu, but when I > uninstalled that virtualenv and reinstalled (with the same > requirements.txt), it no longer can run on gpu. > > > Theano was installed via pip to a virtualenv (managed by pyenv): > pip install numpy > pip install scipy > pip install theano > > For what its worth, I have tried uninstalling and reinstalling many times, > as well as installing from bleeding edge. > > Please consider my situation with theano 8.2, rather than suggesting theano > 9/dev. > > *System/config specs* > > > *.theanorc*[cuda] > root = /usr/local/cuda-7.5 > > [global] > floatX=float32 > optimizer = None > device = gpu > force_device = True > > [nvcc] > fastmath = True > > > *.bashrc*# Source global definitions > if [ -f /etc/bashrc ]; then > . /etc/bashrc > fi > > # User specific aliases and functions > export > LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64" > > export CUDA_HOME=/usr/local/cuda > export PATH=$PATH:/usr/local/cuda/bin > > export PYENV_ROOT="$HOME/.pyenv" > export PATH="$PYENV_ROOT/bin:$PATH" > eval "$(pyenv init -)" > eval "$(pyenv virtualenv-init -)" > > > *which pip*~/.pyenv/shims/pip > > > *pip list*cycler (0.10.0) > h5py (2.6.0) > hickle (2.1.0) > Keras (1.1.0) > matplotlib (1.5.3) > nose (1.3.7) > nose-parameterized (0.5.0) > *numpy (1.11.1)* > pip (8.1.2) > pyparsing (2.1.9) > python-dateutil (2.5.3) > pytz (2016.6.1) > PyYAML (3.12) > *scipy (0.18.1)* > setuptools (27.3.0) > six (1.10.0) > *Theano (0.8.2)* > wheel (0.30.0a0) > > > *nvcc --version* > nvcc: NVIDIA (R) Cuda compiler driver > Copyright (c) 2005-2015 NVIDIA Corporation > Built on Tue_Aug_11_14:27:32_CDT_2015 > Cuda compilation tools, release 7.5, V7.5.17 > > > *gcc --version*gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16) > Copyright (C) 2010 Free Software Foundation, Inc. > This is free software; see the source for copying conditions. There is NO > warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. > > *uname -mrs* > Linux 2.6.32-573.3.1.el6.x86_64 x86_64 > > *lsb_release -a* > LSB Version: > :base-4.0-amd64:base-4.0-noarch:core-4.0-amd64:core-4.0-noarch:graphics-4.0-amd64:graphics-4.0-noarch:printing-4.0-amd64:printing-4.0-noarch > Distributor ID: CentOS > Description: CentOS release 6.7 (Final) > Release: 6.7 > Codename: Final > > > -- > > --- > 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. -- Pascal -- --- 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.
