arcaduf

# To mount the keep
mkdir keep
arv-mount keep/

# To unmount the keep:
arv-mount --unmount keep/
rm -r keep

# To run a workflow
arvados-cwl-runner --project-uuid <PROJECT_UUID> --submit --no-wait file.cwl file.yml

arcaduf

virtualenv ~/mypio
source ~/mypio/bin/activate
pip install scipy scikit-image scikit-learn SimpleITK h5py
deactivate

sbatch shellwrapper.sh

Inside the shellwrapper:
#!/bin/bash
#SBATCH --partition gpu

arcaduf

srun --partition gpu -I --pty /bin/bash -l
ml python2
ml virtualenv
ml CUDA/8.0.44
ml cuDNN/5.1-CUDA-8.0.44
ml OpenCV/2.4.11-goolf-1.7.20
ml TensorFlow/1.1.0-goolf-1.7.20-CUDA-8.0.44-Python-2.7.11

# Check whether tensorflow-gpu works fine:
python

arcaduf

ml python2
ml virtualenv
source ~/mypio/bin/activate
(mypio) pip install jupyter
(mypio) python -m ipykernel install --user --name=mypio
(mypio) jupyter notebook --no-browser --NotebookApp.token='' --NotebookApp.ip='0.0.0.0'

arcaduf

Go to "Device" --> "Install Guest Additions"
As root, run the program VBoxLinuxAdditions.run
Go to "Device" --> "Shared folders" and select the folders of interest
sudo mkdir /media/windows-share
sudo mount -t vboxsf folder-name /media/windows-share

arcaduf

Download directly the Git repo of opencv and opencv_contrib;
Type: "cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D WITH_CUDA=ON -D ENABLE_FAST_MATH=1 -D CUDA_FAST_MATH=1 -D WITH_CUBLAS=1  -D INSTALL_PYTHON_EXAMPLES=ON  -D WITH_IPP=OFF -D BUILD_EXAMPLES=ON -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules ..
"make -j 4"
"sudo make install"
"sudo ldconfig"
Even if it runs on 4 cores, building OpenCV requires 4/4.5 h, so equip yourself with patience.

arcaduf

Simply install Theano with: 'sudo pip install Theano'
Run the Python scripts as: 'THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python script_name.py'

arcaduf

Ensure that you have python 2.7
Run: "sudo apt-get install python-pip python-dev python-virtualenv"
Run: "pip install --upgrade tensorflow-gpu"
Afterwards, if you run python on terminal, you should be able to import tensorflow with an error.
Add this line to .bashrc: "export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64".
Download cudnn version 5 for linux from the official website and copy the files into /usr/local/cuda/include and /usr/local/cuda/lib64 according to how the files are placed in cudnn.

arcaduf

Chech whether your gcc and g++ are the right version, i.e. version 5.
Go to "Software and Updates", then "Additional Drivers"; select NVIDIA drivers 375; install and restart.
Download the CUDA toolkit 8.0 from the official website.
Run the .run file accepting everything except the installation of the NVIDIA drivers, which already exist.
Go to the sample folder; run "make" and try the "matrixMul" script.

arcaduf

Download textext package
Unpack it and copy the .inc and .py file into ~/.config/inkscape/extensions/
Type: "sudo chmod 777 ~/.config/inkscape/extensions/textext.inx"
Type: "sudo chmod 777 ~/.config/inkscape/extensions/textext.inc"