Run Python scripts

DPDispatcher can be used to directly run a single Python script:

dpdisp run script.py

The script must include inline script metadata compliant with PEP 723. An example of the script is shown below.

# /// script
# # dpdispatcher doesn't use `requires-python` and `dependencies`
# requires-python = ">=3"
# dependencies = [
# ]
# [tool.dpdispatcher]
# work_base = "./"
# forward_common_files=[]
# backward_common_files=[]
# [tool.dpdispatcher.machine]
# batch_type = "Shell"
# local_root = "./"
# context_type = "LazyLocalContext"
# [tool.dpdispatcher.resources]
# number_node = 1
# cpu_per_node = 1
# gpu_per_node = 0
# group_size = 0
# [[tool.dpdispatcher.task_list]]
# # no need to contain the script filename
# # `uv run` supports dependencies declared in the PEP-723 format
# # See: https://docs.astral.sh/uv/guides/scripts/
# command = "uv run"
# # can be a glob pattern
# task_work_path = "./"
# forward_files = []
# backward_files = ["log"]
# ///

print("hello world!")

The PEP 723 metadata entries for tool.dpdispatcher are defined as follows:

pep723:

type: dict

argument path: pep723

PEP 723 metadata

work_base:

type: str, optional, default: ./

argument path: pep723/work_base

Base directory for the work

forward_common_files:

type: typing.List[str], optional, default: []

argument path: pep723/forward_common_files

Common files to forward to the remote machine

backward_common_files:

type: typing.List[str], optional, default: []

argument path: pep723/backward_common_files

Common files to backward from the remote machine

machine:

type: dict

argument path: pep723/machine

Machine configuration. See related documentation for details.

batch_type:

type: str

argument path: pep723/machine/batch_type

The batch job system type. Option: Torque, PBS, OpenAPI, DistributedShell, LSF, Slurm, SGE, Fugaku, SlurmJobArray, Bohrium, Shell, JH_UniScheduler

local_root:

type: str | NoneType

argument path: pep723/machine/local_root

The dir where the tasks and relating files locate. Typically the project dir.

remote_root:

type: str | NoneType, optional

argument path: pep723/machine/remote_root

The dir where the tasks are executed on the remote machine. Only needed when context is not lazy-local.

clean_asynchronously:

type: bool, optional, default: False

argument path: pep723/machine/clean_asynchronously

Clean the remote directory asynchronously after the job finishes.

retry_count:

type: int, optional, default: 3

argument path: pep723/machine/retry_count

Number of retries to resubmit failed jobs.

Depending on the value of context_type, different sub args are accepted.

context_type:

type: str (flag key)

argument path: pep723/machine/context_type

possible choices: |code:pep723/machine[OpenAPIContext]|_, |code:pep723/machine[LocalContext]|_, |code:pep723/machine[HDFSContext]|_, |code:pep723/machine[SSHContext]|_, |code:pep723/machine[LazyLocalContext]|_, |code:pep723/machine[BohriumContext]|_

The connection used to remote machine. Option: LocalContext, LazyLocalContext, SSHContext, BohriumContext, HDFSContext, OpenAPIContext

When |flag:pep723/machine/context_type|_ is set to OpenAPIContext (or its aliases openapicontext, OpenAPI, openapi):

remote_profile:

type: dict, optional

argument path: pep723/machine[OpenAPIContext]/remote_profile

The information used to maintain the connection with remote machine. This field is empty for this context.

When |flag:pep723/machine/context_type|_ is set to LocalContext (or its aliases localcontext, Local, local):

remote_profile:

type: dict, optional

argument path: pep723/machine[LocalContext]/remote_profile

The information used to maintain the local machine.

symlink:

type: bool, optional, default: True

argument path: pep723/machine[LocalContext]/remote_profile/symlink

Whether to use symbolic links to replace copy. This option should be turned off if the local directory is not accessible on the Batch system.

When |flag:pep723/machine/context_type|_ is set to HDFSContext (or its aliases hdfscontext, HDFS, hdfs):

remote_profile:

type: dict, optional

argument path: pep723/machine[HDFSContext]/remote_profile

The information used to maintain the connection with remote machine. This field is empty for this context.

When |flag:pep723/machine/context_type|_ is set to SSHContext (or its aliases sshcontext, SSH, ssh):

remote_profile:

type: dict

argument path: pep723/machine[SSHContext]/remote_profile

The information used to maintain the connection with remote machine.

hostname:

type: str

argument path: pep723/machine[SSHContext]/remote_profile/hostname

hostname or ip of ssh connection.

username:

type: str

argument path: pep723/machine[SSHContext]/remote_profile/username

username of target linux system

password:

type: str, optional

argument path: pep723/machine[SSHContext]/remote_profile/password

(deprecated) password of linux system. Please use SSH keys instead to improve security.

port:

type: int, optional, default: 22

argument path: pep723/machine[SSHContext]/remote_profile/port

ssh connection port.

key_filename:

type: str | NoneType, optional, default: None

argument path: pep723/machine[SSHContext]/remote_profile/key_filename

key filename used by ssh connection. If left None, find key in ~/.ssh or use password for login

passphrase:

type: str | NoneType, optional, default: None

argument path: pep723/machine[SSHContext]/remote_profile/passphrase

passphrase of key used by ssh connection

timeout:

type: int, optional, default: 10

argument path: pep723/machine[SSHContext]/remote_profile/timeout

timeout of ssh connection

totp_secret:

type: str | NoneType, optional, default: None

argument path: pep723/machine[SSHContext]/remote_profile/totp_secret

Time-based one time password secret. It should be a base32-encoded string extracted from the 2D code.

tar_compress:

type: bool, optional, default: True

argument path: pep723/machine[SSHContext]/remote_profile/tar_compress

The archive will be compressed in upload and download if it is True. If not, compression will be skipped.

look_for_keys:

type: bool, optional, default: True

argument path: pep723/machine[SSHContext]/remote_profile/look_for_keys

enable searching for discoverable private key files in ~/.ssh/

execute_command:

type: str | NoneType, optional, default: None

argument path: pep723/machine[SSHContext]/remote_profile/execute_command

execute command after ssh connection is established.

proxy_command:

type: str | NoneType, optional, default: None

argument path: pep723/machine[SSHContext]/remote_profile/proxy_command

ProxyCommand to use for SSH connection through intermediate servers.

When |flag:pep723/machine/context_type|_ is set to LazyLocalContext (or its aliases lazylocalcontext, LazyLocal, lazylocal):

remote_profile:

type: dict, optional

argument path: pep723/machine[LazyLocalContext]/remote_profile

The information used to maintain the connection with remote machine. This field is empty for this context.

When |flag:pep723/machine/context_type|_ is set to BohriumContext (or its aliases bohriumcontext, Bohrium, bohrium, DpCloudServerContext, dpcloudservercontext, DpCloudServer, dpcloudserver, LebesgueContext, lebesguecontext, Lebesgue, lebesgue):

remote_profile:

type: dict

argument path: pep723/machine[BohriumContext]/remote_profile

The information used to maintain the connection with remote machine.

email:

type: str, optional

argument path: pep723/machine[BohriumContext]/remote_profile/email

Email

password:

type: str, optional

argument path: pep723/machine[BohriumContext]/remote_profile/password

Password

program_id:

type: int, alias: project_id

argument path: pep723/machine[BohriumContext]/remote_profile/program_id

Program ID

retry_count:

type: NoneType | int, optional, default: 2

argument path: pep723/machine[BohriumContext]/remote_profile/retry_count

The retry count when a job is terminated

ignore_exit_code:

type: bool, optional, default: True

argument path: pep723/machine[BohriumContext]/remote_profile/ignore_exit_code

The job state will be marked as finished if the exit code is non-zero when set to True. Otherwise,

the job state will be designated as terminated.

keep_backup:

type: bool, optional

argument path: pep723/machine[BohriumContext]/remote_profile/keep_backup

keep download and upload zip

input_data:

type: dict

argument path: pep723/machine[BohriumContext]/remote_profile/input_data

Configuration of job

resources:

type: dict

argument path: pep723/resources

Resources configuration. See related documentation for details.

number_node:

type: int, optional, default: 1

argument path: pep723/resources/number_node

The number of nodes required for each job.

cpu_per_node:

type: int, optional, default: 1

argument path: pep723/resources/cpu_per_node

CPU numbers of each node assigned to each job.

gpu_per_node:

type: int, optional, default: 0

argument path: pep723/resources/gpu_per_node

GPU numbers of each node assigned to each job.

queue_name:

type: str, optional, default: (empty string)

argument path: pep723/resources/queue_name

The queue name of batch job scheduler system.

group_size:

type: int

argument path: pep723/resources/group_size

The number of tasks in a job. 0 means infinity.

custom_flags:

type: typing.List[str], optional

argument path: pep723/resources/custom_flags

The extra lines pass to job submitting script header

strategy:

type: dict, optional

argument path: pep723/resources/strategy

strategies we use to generation job submitting scripts.

if_cuda_multi_devices:

type: bool, optional, default: False

argument path: pep723/resources/strategy/if_cuda_multi_devices

If there are multiple nvidia GPUS on the node, and we want to assign the tasks to different GPUS.If true, dpdispatcher will manually export environment variable CUDA_VISIBLE_DEVICES to different task.Usually, this option will be used with Task.task_need_resources variable simultaneously.

ratio_unfinished:

type: float, optional, default: 0.0

argument path: pep723/resources/strategy/ratio_unfinished

The ratio of tasks that can be unfinished.

customized_script_header_template_file:

type: str, optional

argument path: pep723/resources/strategy/customized_script_header_template_file

The customized template file to generate job submitting script header, which overrides the default file.

para_deg:

type: int, optional, default: 1

argument path: pep723/resources/para_deg

Decide how many tasks will be run in parallel.

source_list:

type: typing.List[str], optional, default: []

argument path: pep723/resources/source_list

The env file to be sourced before the command execution.

module_purge:

type: bool, optional, default: False

argument path: pep723/resources/module_purge

Remove all modules on HPC system before module load (module_list)

module_unload_list:

type: typing.List[str], optional, default: []

argument path: pep723/resources/module_unload_list

The modules to be unloaded on HPC system before submitting jobs

module_list:

type: typing.List[str], optional, default: []

argument path: pep723/resources/module_list

The modules to be loaded on HPC system before submitting jobs

envs:

type: dict, optional, default: {}

argument path: pep723/resources/envs

The environment variables to be exported on before submitting jobs

prepend_script:

type: typing.List[str], optional, default: []

argument path: pep723/resources/prepend_script

Optional script run before jobs submitted.

append_script:

type: typing.List[str], optional, default: []

argument path: pep723/resources/append_script

Optional script run after jobs submitted.

wait_time:

type: float | int, optional, default: 0

argument path: pep723/resources/wait_time

The waitting time in second after a single task submitted

kwargs:

type: dict, optional

argument path: pep723/resources/kwargs

Vary by different machines.

batch_type:

type: str, optional

argument path: pep723/resources/batch_type

Allow this key when strict checking.

task_list:

type: list

argument path: pep723/task_list

List of tasks to execute.

This argument takes a list with each element containing the following:

command:

type: str, optional, default: python

argument path: pep723/task_list/command

Python interpreter or launcher. No need to contain the Python script filename.

task_work_path:

type: str

argument path: pep723/task_list/task_work_path

The dir where the command to be executed. Can be a glob pattern.

forward_files:

type: typing.List[str], optional, default: []

argument path: pep723/task_list/forward_files

The files to be uploaded in task_work_path before the task exectued.

backward_files:

type: typing.List[str], optional, default: []

argument path: pep723/task_list/backward_files

The files to be download to local_root in task_work_path after the task finished

outlog:

type: str | NoneType, optional, default: log

argument path: pep723/task_list/outlog

The out log file name. redirect from stdout

errlog:

type: str | NoneType, optional, default: err

argument path: pep723/task_list/errlog

The err log file name. redirect from stderr

$ref support

dpdisp run supports loading external JSON/YAML snippets via $ref in tool.dpdispatcher metadata. For security reasons, this feature is disabled by default.

Enable explicitly with:

dpdisp run script.py --allow-ref