Data Models

HpcConfig

Property	Type	Description	Required	Default
hpc_type	Any	Type of HPC queueing system (such as ‘slurm’)	True
job_prefix	string	Prefix added to each HPC job name	False	job
hpc	Union[SlurmConfig, LocalHpcConfig, FakeHpcConfig]	Interface-specific config options	True

SlurmConfig

Property	Type	Description	Required	Default
account	string	Project account to use	True
partition	string	HPC partition on which to submit	False
reservation	string	HPC reservation on which to submit	False
qos	string	Set to high to get faster node allocations at twice the cost	False
walltime	string	Maximum time allocated to each node	False	4:00:00
gpu gres	string	Request nodes that have at least this number of GPUs. Ex: ‘gpu:2’	False
mem	string	Request nodes that have at least this amount of memory	False
tmp	string	Request nodes that have at least this amount of storage scratch space	False
nodes	integer	Number of nodes to use for each job	False
ntasks	integer	Number of tasks per job (nodes is not required if this is provided)	False
ntasks_per_node	integer	Number of tasks per job (max in number of CPUs)	False

SubmitterParams

This object can be automatically generated with this command:

$ jade config submitter-params

Property	Type	Description	Required	Default
Generate Reports generate_reports	boolean	Controls whether to generate reports after completion	False	True
Hpc Config hpc_config	HpcConfig	HPC config options	True
Max Nodes max_nodes	integer	Max number of compute nodes to use simultaneously, default is unbounded	False
Num Processes num_processes	integer	Number of processes to run in parallel on each node	False
Per Node Batch Size per_node_batch_size	integer	How many jobs to assign to each node	False	500
Node Setup Script node_setup_script	string	Script to run on each node before starting jobs	False
Node Shutdown Script node_shutdown_script	string	Script to run on each node after completing jobs	False
Poll Interval poll_interval	integer	Interval in seconds on which to poll jobs for status	False	10
Resource Monitor Interval resource_monitor_interval	integer	Interval in seconds on which to collect resource stats. Disable monitoring by setting this to None/null.summaries of stats.	False	10
resource_monitor_type	Any	Type of resource monitoring to perform. Options: [‘aggregation’, ‘periodic’, ‘none’]	False	aggregation
Resource Monitor Stats resource_monitor_stats	Any	Resource utilization stats to monitor	False	{‘cpu’: True, ‘disk’: False, ‘memory’: True, ‘network’: False, ‘process’: False, ‘include_child_processes’: True, ‘recurse_child_processes’: False}
Try Add Blocked Jobs try_add_blocked_jobs	boolean	Add blocked jobs to a batch if all blocking jobs are in the batch. Be aware of time constraints.	False	True
Time Based Batching time_based_batching	boolean	Use time-based batching instead of job-count-based batching	False	False
Dry Run dry_run	boolean	Dry run mode; don’t start any jobs	False	False
Verbose verbose	boolean	Enable debug logging	False	False
Singularity Params singularity_params	Any	Singularity container parameters	False
Distributed Submitter distributed_submitter	boolean	Submit new jobs and update status on compute nodes.	False	True

SubmissionGroup

Property	Type	Description	Required	Default
name	string	User-defined name of the group	True
submitter_params	SubmitterParams	Submission parameters for the group	True