Installation

Custom

Conda Installation

  1. Download Miniconda Installer

    • Visit the Miniconda Downloads page.

    • Select the Linux installer for Python 2.x or 3.x as per your requirements.

  2. Open a Terminal

    • Open a terminal window on Linux by pressing Ctrl + Alt + T, or search for “Terminal” in the applications menu.

  3. Navigate to Download Directory

    cd ~/Downloads

    Change to the directory where the installer was downloaded, usually the Downloads directory.

  4. Make the Installer Executable

    • Make the downloaded script executable. Replace Miniconda3-latest-Linux-x86_64.sh with the actual downloaded file name.

      chmod +x Miniconda3-latest-Linux-x86_64.sh

  5. Run the Installer

    • Execute the installer script and follow the on-screen instructions.

      ./Miniconda3-latest-Linux-x86_64.sh

    • You’ll need to approve the license agreement and choose the installation location.

  6. Initialize Conda

    • After installation, initialize Miniconda to add Conda to your PATH.

  7. Close and Reopen Your Terminal

    • To apply the changes, close and reopen your terminal window.

Creating Environment

Create python 3.8 environment

Note

The name of the conda environment in this example is tiegcm.

conda create --name tiegcm python=3.8

Installing gcmprocpy

Warning

cartopy requires geos which doesn’t install properly via the pip install. Use the command below if you face the issue.

conda install -c conda-forge cartopy

To install gcmprocpy, run the following command:

pip install gcmprocpy

NCAR Derecho

Creating Environment

Load Conda module

module load conda

Create python 3.8 environment

Note

The name of the conda environment in this example is tiegcm.

conda create --name tiegcm python=3.8

Activate Environment

Note

Make sure the conda module is loaded.

conda activate tiegcm

Installing gcmprocpy

Warning

cartopy requires geos which doesn’t install properly via the pip install. Use the command below if you face the issue.

conda install -c conda-forge cartopy

To install gcmprocpy, run the following command:

pip install gcmprocpy

Installing gcmprocpy for Jupyter Notebooks

Note

Make sure the conda module is loaded.

conda activate tiegcm

Install ipykernal to use the conda environment for Jupyter notebooks.

pip install ipykernel

VS Code Jupyter Notebooks (Casper Nodes)

### Step 1: Request an Interactive Session

To begin, you need to request an interactive session on a compute node using qsub. This will allocate resources for your job, allowing you to use multiple processors instead of running on the login node.

  1. Open your terminal on the Casper login node and enter the following command to request an interactive session:

    `bash qsub -I -A P28100045 -q casper -l select=1:ncpus=4:mpiprocs=4 -l walltime=01:00:00 `

    • -I specifies an interactive session.

    • -A P28100045 is your project/account code (replace with your own).

    • -q casper requests the Casper queue.

    • -l select=1:ncpus=4:mpiprocs=4 requests 1 node with 4 CPUs and 4 MPI processes.

    • -l walltime=01:00:00 specifies the walltime limit (1 hour in this case).

  2. Once the job is submitted, you will see the following output indicating the job’s status:

    `bash qsub: waiting for job 2884283.casper-pbs to start qsub: job 2884283.casper-pbs ready `

### Step 2: Check the Hostname of the Compute Node

After the job is ready, you need to check the hostname of the compute node that has been allocated to you.

  1. Run the following command to display the hostname of your current session:

    `bash echo $HOSTNAME `

    This will output something like:

    `bash crhtc62 `

    This hostname is used to connect to the compute node via SSH.

### Step 3: Connect to the Compute Node from VSC

  1. Open Visual Studio Code on your local machine.

  2. Use the Remote-SSH extension in VSC to connect to the compute node.

  3. In VSC, configure the SSH connection to the node using the following format:

    ` $HOSTNAME.hpc.ucar.edu `

    Replace $HOSTNAME with the actual hostname from the previous step (e.g., crhtc62).

  4. Once connected, you can start editing and running code on the compute node as needed.

Note

This process only works on Casper and will not work on Derecho due to firewall rules on Derecho.

NCAR JupyterHub

Warning

NCAR JupyterHub only workes when matplotlib is in inline backend. Use the following at the start of your Jupyter notebook to enable inline backend.

%matplotlib inline

Open JupyterHub by visiting the NCAR JupyterHub.

Create a new Jupyter notebook by clicking on the New button or select an existing notebook.

Change the kernel to the conda environment by clicking on the Kernel tab on the top left and selecting the conda environment. The conda environment will be listed as Python [conda env:tiegcm], where tiegcm is the name of the conda environment.

Follow API documentation to use gcmprocpy in the Jupyter notebook.

NASA Pleiades

Creating Environment

Load Conda module

module use -a /swbuild/analytix/tools/modulefiles
module load miniconda3/v4

Note

Replace $USER with your username on Pleiades.

export CONDA_PKGS_DIRS=/nobackup/$USER/.conda/pkgs

Create python 3.8 environment

conda create -n tiegcm python=3.8

Activate Environment

Note

The name of your environment will be set to my_{environment_name} due to Pleiades deployment. Make sure the conda module is loaded.

conda activate my_tiegcm

Installing gcmprocpy

Warning

cartopy requires geos which doesn’t install properly via the pip install. Use the command below if you face the issue.

conda install -c conda-forge cartopy

To install gcmprocpy, run the following command:

pip install gcmprocpy