hostname

nvidia-smi

# Container starting directory determined by settings in Docker image

# For NVIDIA NGC containers this is normally /workspace

echo Container starts in the directory:

pwd

# Install required packages

# Set environment variable to install packages to world-writable location inside container

export PYTHONUSERBASE=/workspace/.local

# Escaping PYTHONUSERBASE in the next line so that it is expanded inside docker and take the value set inside the container

mkdir -p $PYTHONUSERBASE/bin

export PATH=$PATH:$PYTHONUSERBASE/bin

echo PATH inside container

echo PATH=$PATH

pip install -U -q --no-cache-dir --user Braindecode==0.4.85

# change directory to correct location if required

cd /home/users/ntu/kzhang01/hbm/eeg-adapt

echo Changed working directory to

pwd

datapath=/home/users/ntu/kzhang01/scratch/KU_mi_smt.h5

modelpath=/home/users/ntu/kzhang01/hbm/eeg-adapt/result_base/best

logpath=/home/users/ntu/kzhang01/hbm/eeg-adapt/tasks_adapt{0}

# Fix file locking disabled on this file system (/scratch) when reading hdf5

export HDF5_USE_FILE_LOCKING='FALSE'