Interactive Apps

The Interactive Apps dropdown from the toolbar will list a few standalone programs you are able to launch directly from the browser as well as an HPC Desktop that will allow you access all of the other software on Cheaha.

Currently, the available standalone programs are IGV, Matlab, RStudio, SAS, and Jupyter.

All of the interactive apps have similar setup pages. For instance, if we click HPC Desktop, the following screen will appear:

This will allow to choose the number of hours, partition, number of cpus, and memory per cpu needed for the job. These fields are common to all interactive apps and are required. Not all partitions are available when creating an interactive job in OOD. For instance, if you need to use the largemem partition, request those resources in a terminal session for an interactive job or submit a batch job.

Tip

You can decrease wait time in the queue by choosing resources carefully. The closer your request is to actual usage, the more optimal your wait time will be. Please see our section on Job Efficiency for more information.

Once you've selected the compute resources you need, Launch the job. This will bring you to the My Interactive Sessions page, and read on.

My Interactive Sessions

The My Interactive Sessions page looks like:

For each job running via Open OnDemand, there will be a card listed on this page. Each card has basic information about the number of cores, nodes, and time remaining for the job. Also shown are the Job ID in the top-left part of the job card, and a Session ID link near the center of the card. The Job ID and Session ID are important for diagnosing issues you may encounter on Cheaha while using Open OnDemand. The Session ID link points to a folder in your home directory with information we may ask for.

Click the Launch Desktop in new tab button to open your interactive VNC session. You may have to wait for the card to change from blue to green to launch the job.

Bug

If your job fails to launch, please see our FAQ for possible solutions, or contact us.

Note

For HPC Desktop, you do not need to request resources after you open the Desktop. You are already on a compute node. Any tasks you run will use the resources you requested when initializing the job.

Note

You can request another interactive session in a terminal in HPC Desktop. Only the terminal you requested the other interactive session in will have access to the new resources. Everything else in the HPC Desktop will run with the resources you requested when creating the initial job.

These interactive jobs can be stopped early by clicking Delete on the right side of the job card.

Standalone Programs

As shown earlier, some software can be run outside of the VNC session. Setup for most of these follow the same rules as creation of an HPC Desktop job in terms of requesting resources. You will also need to select the version of software to use for the job.

Note

Versions shown in the OOD form may not line up with versions available in modules. If you need a version not available in OOD, please feel free to send in a ticket.

RStudio Server

RStudio is available for use graphically in your browser via OOD. As with other standalone programs, you'll need to select the resources required using the job creation form. You'll also need to select both the version of RStudio you wish to use, and the version of R you wish to use. To adjust the environment, please use the Environment Setup field to load modules besides R and RStudio as seen below. All other modules and paths should be loaded here as it is difficult to load and consistently use modules once RStudio starts.

Important

Unless an older version of R is absolutely necessary, it is highly suggested to always use the newest version of R and RStudio for both updated functionality within those software as well as updated compilers for package installation. Using the newest version of R solves most known package installation errors.

RStudio and Python

If you have a workflow that uses both R and Python, it is strongly recommended to use the reticulate package along with Anaconda environments. Reticulate allows researchers to load Python packages into a native R session as objects. For instance, if someone prefer some functionality of the pandas package but has other code already written in R, they can import pandas to R and use both simultaneously.

This also allows researchers to download precompiled command line binaries into an Anaconda environment and easliy use them in their R scripts.

For setup, use the following steps:

1. In a terminal on a compute node, either in an HPC Desktop job or by clicking the blue Host button on any job card:

   1. Load the Anaconda3 module
   2. Create an Anaconda environment. More information about how to create Anaconda environments can be found in our documentation.
   3. Activate your environment and install your requuired python packages using either pip install or conda install depending on the package source.

   Note

   The preceding steps should only need to be run once. If other Python packages need to be installed in the same environment, repeat steps 1 and 3. You will not need to recreate your environment.

2. In RStudio:

   1. Add the command module load Anaconda3 to the Environment Setup window when requesting the RStudio job.
   2. If not already installed, install the reticulate package using either install.packages or the renv package.
   3. Use reticulate::use_condaenv('env_name') to load your conda environment.
   4. From here, you will be able to interact with all of the python packages and non-python precompiled binaries in your Anaconda environment using R and RStudio. Please read more about how to do that in reticulate's documentation.

For cases where your R code only needs access to precompiled binaries or libraries and does not need to import any Python libraries, you can instead create your Anaconda environment and add the following lines into the Environment Setup window:

module load Anaconda3
conda activate <env_name>

This will add those binaries and libraries to your environment $PATH which RStudio will inherit.

Important

If you're wanting to directly use any Python package in R, DO NOT include the conda activate command in the Environment Setup. Use reticulate instead as described above.

RStudio Projects and renv

The most recent versions of RStudio installed on Cheaha support R Projects as well as package management through the renv package. Please read more about improving analysis reproducibility using both of these tools in our workflow solutions

Using Pandoc and knitr within RStudio

If you want to use RMarkdown to create reports in RStudio, R modules using version 4.2.0 and later include knitr compatibility. Please use the latest versions of both R and Rstudio for fully integrated knitr functionality.

Starting With a Clean Session to Avoid Errors

By default, RStudio loads the most recently opened project at startup and restores the .RData file into the workspace. If you only work on a single project, this may be helpful. If you frequently change projects then these default settings can create difficult-to-diagnose errors, or you may inadvertently alter a project by adding incorrect packages, for example.

To reduce the risk of these kinds of errors, uncheck the highlighted fields below in the RStudio Options menu under the "General" selection.

• Restore most recently opened project at startup
• Restore .RData into workspace at startup

Jupyter Notebook

Jupyter Notebooks are available for use graphically in your browser via OOD. As with other standalone programs, you'll need to select the resources required using the job creation form. The form is shown below.

Jupyter Notebooks are commonly used with Anaconda environments. If you are unfamiliar with Anaconda environments please see the Working with Anaconda Environments section below before continuing here.

To modify the Operating System (OS) environment that Anaconda and Jupyter will run in, please use the Environment Setup field to load modules. For GPU applications you'll need to load a CUDA/* module. If working with deep learning workflows, you will also possibly need to load the cuDNN/*-CUDA-* module corresponding to your choice of CUDA/* module version. These are required for popular ML/DL/AI libraries like TensorFlow, Keras, and PyTorch. Use module spider cuda and module spider cudnn to view the list of appropriate modules. An example of what to put in the Environment Setup field, when using Tensorflow in a Jupyter notebook, is shown below.

# ENVIRONMENT SETUP
module load CUDA/12.2.0
module load cuDNN/8.9.2.26-CUDA-12.2.0

For information on which versions of CUDA to load for Tensorflow and PyTorch, please see Tensorflow Compatibility and PyTorch Compatibility.

Note

If you get a Failed to Connect message when opening the job, close the tab and wait a couple of minutes. Jupyter is still initializing and takes some time after the job first begins running.

Important

If you are not able to see your environment, you may need to install the ipykernel package. It is required for Jupyter to recognize your environment. See Packages for Jupyter for more information.

Important

Do not load module load Anaconda3 in the Environment Setup field, as it is loaded automatically. Loading any versions of Anaconda3 would affect the Python executable, which is used by default. These results in hard-to-diagnose errors in the OOD Jupyter notebook.

Warning

Having conda/mamba activate and source activate statements in the Environment Setup field can cause unexpected and silent job failure. Avoid using conda activate in the Environment Setup field.

Working with Anaconda Environments

For information on working with Anaconda environments please see our Using Anaconda page. The please review our Cheaha-specific Anaconda page for important tips and how to avoid common pitfalls.

Extra Jupyter Arguments

The Extra Jupyter Arguments field allows you to pass additional arguments to the Jupyter Server as it is being started. It can be helpful to point the server to the folder containing your notebook. To do this, assuming your notebooks are stored in /data/user/$USER, also known as $USER_DATA, put --notebook-dir=$USER_DATA in this field. You will be able to navigate to the notebook if it is in a subdirectory of notebook-dir, but you won't be able to navigate to any other directories. An example is shown below.

Submitting the Jupyter Notebook Job

Submitting the job will bring you to the My Interactive Jobs window while the Jupyter job is initialized. Click Connect to Jupyter to open the Jupyter Home Page.

Note

If you get a Failed to Connect message when opening the job, close the tab and wait a couple of minutes. Jupyter is still initializing and takes some time after the job first begins running.

The Jupyter Server Home Page

The Jupyter Server Home Page will look like the following

From here, you can navigate to and select an existing notebook, or you can create a new one using one of your existing virtual environments or the base environment. Once inside a Jupyter notebook, you can use the Kernel --> Change kernel menu to select your preferred Anaconda environment.

Important

See Anaconda Environments for information on Jupyter related packages.

Warning

Files deleted within using the Jupyter server interface are not truly deleted. Instead they are moved to $HOME/.local/share/Trash. This may cause inflation of personal storage usage on Cheaha.

Python Libraries and Virtual Environments

To run Jupyter with specific libraries and packages outside of the base install, you will need to create a virtual environment first. You can do this either in an HPC Desktop job or in the Conda tab of the Jupyter homepage.

The Conda has the following layout:

1. Current environments (red): a listing of the current existing environments in your $HOME/.conda/envs folder.
2. Available packages (green): a list of all packages available to install from conda sources.
3. Installed packages (blue): a list of the packages installed in the currently selected environment.

To create a new environment, click the + button at the top of the Current environments pane and enter the name of the environment. After it has been created, you can select packages to install by searching for the package name at the top right of the Available packages pane. After selecting the package, click the -> button, and the package and all its dependencies will be installed.

Note

If a package is not available using the conda command directly, it will not be listed as an available package. Use a terminal window to install the package as necessary.

Note

In order to use an environment with Jupyter, the ipykernel library is necessary. Creating an environment in the Conda tab will autoinstall this library. If using the terminal, use conda install ipykernel to install it.

After successfully creating your environment, navigate to the Files tab. You can create a new notebook using the New dropdown menu in the top right. Select your virtual environment of choice, and a notebook will be created and opened.

Help GPU is not Available with TensorFlow or PyTorch

If you are using Jupyter with TensorFlow or PyTorch and no GPU is found, please see our Slurm GPU page sections on TensorFlow Compatibility and PyTorch Compatibility. For MATLAB, please see MATLAB Compatibility.

Matlab

Matlab is available for use graphically in your browser via OOD. As with other standalone programs, you'll need to select the resources required using the job creation form. The form is shown below.

Warning

Matlab tends to consume substantial memory at startup. You may experience difficulty with job errors below 20 GB of total memory.

Using Anaconda Python from within Matlab

Matlab has the ability to interoperate with Python from within Matlab. The official documentation for this featuer may be found at https://www.mathworks.com/help/matlab/call-python-libraries.html.

This section is dedicated to using this feature with Anaconda on Cheaha. To use Python contained in an Anaconda Environment within Matlab, please use the following steps.

1. Create an HPC Interactive Desktop Job using Open OnDemand.
2. Open a terminal in that job. The following steps should all be run in this terminal unless otherwise specified.
3. Load the Anaconda Module.
4. Create an Environment in Anaconda with the packages needed.
5. Activate the Environment,
6. Load the Matlab Module.
7. Start Matlab by entering the command matlab.

8. Verify success by entering pyenv at the Matlab prompt (not the terminal window). Multiple lines of text will be returned at the prompt. Among them you should see a line like the following, with your environment name in place of <env_name>.

   Executable: /home/$USER/.conda/envs/<env_name>/bin/python
   

You may optionally verify that Python works correctly by entering py.list(["hello", "world"]). A python list object should appear in the workspace.

Using a GPU with MATLAB

Please see the MATLAB Section on our GPU Page.

---

Last update: September 29, 2023