Centre de Calcul Recherche et Technologie with NVIDIA environnement TGCC-CCRT/nvidia

Description

This plugin deals with specific environment characteristics of the cluster at the Très Grand Centre de calcul (France), more specifically the Nvidia GPUs partitions of the Centre de Calcul Recherche et Technologie.

YAML arguments

The following arguments are used to configure the plugin. pyCIF will return an exception at the initialization if mandatory arguments are not specified, or if any argument does not fit accepted values or type:

Optional arguments

python : str, optional, default “python -m mpi4py -rc initialize=False”

the python command used to run sub-instances of pyCIF

python_venv : str, optional

path to the python virtual environment to use

python_module : str, optional

the python module to load, by default python3/3.10.6. This argument can not be used together with ‘job_env’.

job_env : “list of str”, optional

List of commands to execute at the start of the job script to setup the environment. Using this argument will override the default module loading. This argument can not be used together with ‘python_module’. The list list of commands should not include any python virtual environment activation command, please use the ‘python_venv’ argument for that purpose.

gpu : bool, optional, default False

change the command used to run parallel programs in order to allocate one GPU

partition : “rome” or “skylake” or “a64fx” or “v100” or “v100l” or “v100l-os” or “hybrid” or “xlarge” or “v100xl”, optional

partition on which to submit job, used as the -q option for the cc_msub command

project : str, optional

project on which to submit job, used as the -A option for the cc_msub command

filesystem : “Any subset of [‘scratch’, ‘work’, ‘store’], separated by commas or ‘all’”, optional, default “all”

the file system(s) required by the job, used as the -m option for the cc_msub command

qos : “long” or “normal” or “test”, optional, default “normal”

Quality of Service (QoS) used to submit job, used as the -Q option for the cc_msub command

walltime : int, optional, default 7200

maximum walltime of the submited job, used as the -T option for the cc_msub command

nodes : int, optional, default 1

number of nodes to use by the jobs launched by the CIF, used as the -n option for the cc_msub command

cores : int, optional, default 1

number of cores to use by the jobs launched by the CIF, used as the -c option for the cc_msub command

submit_msub : bool, optional, default True

Submit the job with ccc_msub. If false simply run it within the same instance

Requirements

The current plugin requires the present plugins to run properly:

Requirement name	Requirement type	Explicit definition	Any valid	Default name	Default version
model	Model	True	True	None	None

YAML template

Please find below a template for a YAML configuration:

platform:
  plugin:
    name: TGCC-CCRT
    version: nvidia
    type: platform

  # Optional arguments
  python: XXXXX  # str
  python_venv: XXXXX  # str
  python_module: XXXXX  # str
  job_env: XXXXX  # list of str
  gpu: XXXXX  # bool
  partition: XXXXX  # rome|skylake|a64fx|v100|v100l|v100l-os|hybrid|xlarge|v100xl
  project: XXXXX  # str
  filesystem: XXXXX  # Any subset of ['scratch', 'work', 'store'], separated by commas or 'all'
  qos: XXXXX  # long|normal|test
  walltime: XXXXX  # int
  nodes: XXXXX  # int
  cores: XXXXX  # int
  submit_msub: XXXXX  # bool