-
Notifications
You must be signed in to change notification settings - Fork 4
Using this repository on HPC resources
You can find more in-depth resources using the official documentation:
The least error-prone way to use python on one of the supercomputing resources is by installing anaconda yourself. I've generally had issues loading specific modules such as numpy and mdtraj. In order to install anaconda, download the linux version of the package from their website. Make sure to get the 64-bit installer.
Copy the downloaded file onto the supercomputer by your method of choice (globus.org, scp, WinSCP etc.).
- Summit: place file somewhere in the /projects/username/ directory.
- Bridges: place file somewhere in the /home/username/ directory.
Note: The location where you place files and run jobs is important on each resource. Take a look at their documentation (Summit, Bridges) for more details on the intended use of each filesystem.
In the directory where you stored the installer file, give the file permission to execute and then run the script in order to install anaconda:
chmod +x Anaconda3-x.x.x-Linux-x86_64.sh && ./Anaconda3-x.x.x-Linux-x86_64.sh
Make sure the install location is somewhere under the same directory where the installer is located. The default should be acceptable.
Anaconda will automatically add the path to the python executable to your PATH in the ~/.bashrc file. (Note that the Summit admins prefer that we update the ~/.my.bashrc instead of the ~/.bashrc file. You'll need to manually add the path to ~/.my.bashrc) Run exec bash then type python in the console. A python shell should appear.
If that doesn't work, then the path to the python executable might not have been added to your .bashrc file. Open up your .bashrc file (vim ~/.bashrc on Bridges, vim ~/.my.bashrc on Summit) and add the following line: export PATH="/projects/username/anaconda3/bin:$PATH"
Finally, you'll need to add a few modules that don't come with anaconda.
There are a few modules that you will need to download in order to be fully compatible with the scripts in this repository. On your local workstation, you can just run sudo pip install module_name, but that doesn't work on a community resource such as Bridges or Summit since the root directory does not belong to any individual users. Each system has their own way of installing python packages:
Summit: See these instructions.
Bridges: Just run pip install --user package_name
For full compatibility with this repository, you'll need to download the following modules:
- mdtraj
- tqdm
- pymbar
The supercomputing resource you are using may have a python distribution already installed. Sometimes this is sufficient for everything in this repository. But usually not. Here is how to use the installed python modules.
Check to see what versions of python are already available:
Summit: First you must ssh onto a compile node: ssh scompile. Next type module spider python or module avail python and you will be presented with a list of available modules. Load the module with module load python/x.x.x.
Bridges: module avail python will list all available versions. Bridges has a lot of python modules and many (most) of them don't have numpy or are compiled wrong. Your best hope is to use python/3.6.4_gcc5_np1.14.5.
Clone this repository into the appropriate directory (/projects/username/ on Summit, /home/username/ on Bridges):
git clone https://github.com/shirtsgroup/LLC_Membranes.git
Add the following lines to your .bashrc (.my.bashrc on Summit):
- export PYTHONPATH="$PYTHONPATH:/projects/username/LLC_Membranes"
- export PATH=$PATH:/projects/username/LLC_Membranes/analysis/
- export PATH=$PATH:/projects/username/LLC_Membranes/setup/
- export PATH=$PATH:/projects/username/LLC_Membranes/llclib/
- export PATH=$PATH:/projects/username/LLC_Membranes/
Then execute exec bash.
Please refer here for instructions on installing GROMACS on HPC resources.
Please see (COMING SOON) for more information on creating batch submission scripts. Or use one of the premade scripts located here.