radio_transients

Overview

These are my Singularity Recipes for common radio transient software. There are three containers

radio_transients

Contains everything (CPU+GPU)

CUDA 10.2
FETCH          https://github.com/devanshkv/fetch  -- In Conda environment `FE`
heimdall       https://sourceforge.net/p/heimdall-astro/wiki/Use/
- dedisp       https://github.com/ajameson/dedisp
htop           https://htop.dev/
iqrm_apollo    https://gitlab.com/kmrajwade/iqrm_apollo
jupyterlab     https://jupyter.org/
PRESTO         https://www.cv.nrao.edu/~sransom/presto/
psrdada        http://psrdada.sourceforge.net/
psrdada-python https://github.com/TRASAL/psrdada-python
psrcat         https://www.atnf.csiro.au/people/pulsar/psrcat/download.html
pysigproc      https://github.com/devanshkv/pysigproc
riptide        https://github.com/v-morello/riptide
sigproc        https://github.com/SixByNine/sigproc
Tempo          http://tempo.sourceforge.net/
RFIClean       https://github.com/ymaan4/RFIClean
YAPP           https://github.com/jayanthc/yapp
your           https://github.com/thepetabyteproject/your

Get with singularity pull --arch amd64 library://josephwkania/radio_transients/radio_transients:latest

*One of FETCH's dependencies causes PRESTO's Python scripts to fail. This necessitated putting them in different environments. Everything except for PRESTO is in RT, which is loaded by default. PRESTO is in PE, in the shell you can activate this with conda activate PE. If you need access outside the container, you should use radio_transients:cpu, which has PRESTO in the default environment.

radio_transients_cpu

Contains CPU based programs

htop
iqrm_apollo
jupyterlab   
PRESTO
psrcat
pysigproc
riptide
sigproc
Tempo 
RFIClean
YAPP  
your

Get with singularity pull --arch amd64 library://josephwkania/radio_transients/radio_transients:cpu
There is an arm version Singularity.arm, singularity pull --arch arm library://josephwkania/radio_transients/radio_transients:arm

radio_transients_gpu

Contains gpu based programs

CUDA 10.2
FETCH      
jupyterlab
heimdall
- dedisp
htop 
psrdada 
psrdada-python
your

Get with singularity pull --arch amd64 library://josephwkania/radio_transients/radio_transients:gpu

How to use

Your $HOME automatically gets mounted. You can mount a directory with -B /dir/on/host:/mnt, which will mount /dir/on/host to /mnt in the container.

For the gpu processes, you must pass --nv when running singularity.

singularity shell --nv -B /data:/mnt radio_transients_gpu.sif will mount /data to /mnt, give you GPU access, and drop you into the interactive shell.

singularity exec --nv -B /data:/mnt radio_transients_gpu.sif your_heimdall.py -f /mnt/data.fil will mount /data to /mnt, give you GPU access, and run your_heimdall.py without entering the container.

All the Python scripts are installed in a Conda environment RT, this environment is automatically loaded.

You can see the commits and corresponding dates by running singularity inspect radio_transients.sif

Sylabs Cloud

These are built on a E5 v3 family machine and uploaded to Sylabs Cloud at https://cloud.sylabs.io/library/josephwkania/radio_transients/radio_transients They where last built on 27-Nov-2021

If your processor your processor is significantly older than this, you may run into problems with the older processor not having the whole instruction set needed. In this case, you should build use singularity to build the image locally.

An archival version of these (built 25-April-2021) are on Singularity Hub at: https://singularity-hub.org/collections/5231

Improvements

If you come across bug or have suggestions for improvements, let me know or submit a pull request.

Thanks

To Kshitij Aggarwal for bug reports and suggestions.