neuralhydrology_on_hpc

# Get on a compute node (from login node via putty or via interactive shell with ondemand)
srun --account=lp_ees_swm_ls_001 --nodes=1 --time=05:00:00 --ntasks-per-node=1 --gpus-per-node=0 --partition=interactive --clusters=wice --x11 --pty bash -l

# Install miniconda
wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh

# Open miniconda.sh with vim
# Change to your vsc number
# PREFIX="/data/leuven/317/vsc31786/miniconda3-wice
#

# Make it an executable
chmod a+x miniconda3-wice.sh

# Execute command and follow instructions
./miniconda3-wice.sh
# Update conda environment
conda update -n base -c defaults conda

# get yml files
cd /data/leuven/317/vsc31786/miniconda3-wice
rsync -av /staging/leuven/stg_00024/OUTPUT/michelb/miniconda3-wice/*yml
# install conda environment using yml file
conda env create -f environment_cpu.yml
# or for gpu environment
conda env create -f environment_cuda11_8.yml

#Activate new environment
source activate neuralhydrology-cpu
conda activate neuralhydrology-cpu
# or
conda activate neuralhydrology-gpu

# for pytorch with CUDA support (i.e. for neuralhydrology-gpu) run also
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# add neuralhydrology to this environment
cd $VSC_DATA

git clone https://github.com/neuralhydrology/neuralhydrology.git
cd  neuralhydrology
pip install -e .

python -m ipykernel install  --prefix=${VSC_HOME}/.local/ --name neuralhydrology-cpu

### Now you are set for the first example