Running on Clusters / in Parallel¶

snmachine on NERSC¶

To run snmachine on NERSC first open an ssh connection to NERSC, set your shell to bash shell, and run the following from the command line:

export PATH=/global/projecta/projectdirs/lsst/groups/SN/miniconda3/bin:$PATH

Then to use snmachine any time all that is needed is to activate the environment using,

module load python
source activate snmachine

Setting up snmachine on Jupyter @ NERSC¶

Prior to accessing the Jupyter interface, either create a .condarc file (if you do not have already) with the following lines:

envs_dirs:
- /global/projecta/projectdirs/lsst/groups/SN/miniconda3/envs

or append this to your existing .condarc file. Be sure to access Jupyter via the Jupyter-Dev hub , and not the Jupyter-Hub via ‘MyNERSC’ pages. Then, when prompted or accessing the notebook file, choose the Python [conda env:snmachine] kernel. This can be found under Kernel tab in the top menu bar. Now snmachine will be accessible through your Jupyter notebook.

Running jobs @ NERSC¶

This will be implmented in the future.

snmachine on Hypatia¶

First, login to Hypatia with your credentials via the ssh method.

ssh USERNAME@hypatia-login.hpc.phys.ucl.ac.uk

Next you will need to import python into your node using

module load python

Then, if you have not already setup snmachine, you will need to clone the snmachine repository,

git clone https://github.com/LSSTDESC/snmachine.git

and install the package as usual, following steps 2 through 4 in the conda installation guide, Using Conda. You should now be setup to run snmachine as normal on Hypatia.

Run pipeline instructions (temporary)¶

These instructions are a temporary (as of July 2nd 2018) guide to run the jobs pipeline on Hypatia at UCL.

With a working installation of snmachine you will first need to switch to the issue/38/refactor-utils-cjonly branch and pull the latest commits.
```
git pull
git checkout issue/38/refactor-utils-cjonly
```
Navigate to the snmachine/utils folder and execute,
```
python create_jobs.py
```

NOTE: If this fails it is possible that the create_jobs.py script did not find the data files specified by default. In order to try again, first, remove the snmachine/utils/jobs folder. Then, you can then pass your own list of data through the -op argument flag. In doing so you should specify the location to a *.LIST file that contains a list of all the objects that you want to run. An example of this is,
```
python create_jobs.py -op /share/hypatia/snmachine_resources/data/DES_spcc/SIMGEN_PUBLIC_DES/SIMGEN_PUBLIC_DES.LIST
```

Next navigate to the newly created snmachine/utils/jobs folder and execute:
```
sh run_all.sh
```
Your jobs will now be in the queue to execute on the nodes specified in the create_jobs.py script. You can monitor their progress through the showq and qstat shell commands.