pykanto.utils.slurm.launch#

Submit a slurm job that employs distributed/parallel computation using ray.

Functions

submit_job()

Parses arguments and submits a ray job to slurm.
pykanto.utils.slurm.launch.submit_job()[source]#

Parses arguments and submits a ray job to slurm. Code from Peng Zhenghao; modifications (c) 2021 Nilo M. Recalde.

See source code by Peng Zhenghao. Also see ray instructions.

Run pykanto-slaunch --help for arguments. Output bash and log files are saved in a /logs directory within the directory from which you called the script. You can easily change this behaviour by editing the out_dir below

Note

This works as of 2022 @ Oxford University ARC HPC. Chances are it will not work for you ‘out of the box’; this submodule is intended more as a guide or reference than a foolproof way of submitting multi-node / GPU ray jobs.