Skip to content

queue #

Utilities for interacting with HPC queues

Classes:

Functions:

SLURMOptions pydantic-model #

Bases: BaseModel

Options for a Slurm job submission.

Fields:

n_nodes pydantic-field #

n_nodes: int

The number of nodes to request.

n_tasks pydantic-field #

n_tasks: int

The number of tasks to request.

n_cpus_per_task pydantic-field #

n_cpus_per_task: int = 1

The cpus per task.

n_gpus_per_task pydantic-field #

n_gpus_per_task: int = 1

The number of gpus to request per task.

walltime pydantic-field #

walltime: str = '5-0'

The maximum walltime.

partition pydantic-field #

partition: str

The partition to run on.

account pydantic-field #

account: str | None = None

The account to use.

job_name pydantic-field #

job_name: str = 'femto'

The job name to use.

reservation pydantic-field #

reservation: str | None = None

The reservation to use.

model_dump_yaml #

model_dump_yaml(
    output_path: Path | None = None, **kwargs
) -> str

Dump the model to a YAML representation.

Parameters:

  • output_path (Path | None, default: None ) –

    The (optional) path to save the YAML representation to.

Returns:

  • str –

    The YAML representation.

Source code in femto/md/utils/models.py 
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
def model_dump_yaml(self, output_path: pathlib.Path | None = None, **kwargs) -> str:
    """Dump the model to a YAML representation.

    Args:
        output_path: The (optional) path to save the YAML representation to.

    Returns:
        The YAML representation.
    """

    model_yaml = yaml.safe_dump(self.model_dump(), **kwargs)

    if output_path is not None:
        output_path.parent.mkdir(exist_ok=True, parents=True)
        output_path.write_text(model_yaml)

    return model_yaml

to_flags #

to_flags() -> list[str]

Convert this options to a list of sbatch / srun flags.

Source code in femto/fe/utils/queue.py 
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
def to_flags(self) -> list[str]:
    """Convert this options to a list of sbatch / srun flags."""

    return [
        f"--nodes={self.n_nodes}",
        f"--ntasks={self.n_tasks}",
        f"--gpus-per-task={self.n_gpus_per_task}",
        f"--cpus-per-task={self.n_cpus_per_task}",
        f"--partition={self.partition}",
        f"--time={self.walltime}",
        f"--job-name={self.job_name}",
        *([f"--account={self.account}"] if self.account is not None else []),
        *(
            [f"--reservation={self.reservation}"]
            if self.reservation is not None
            else []
        ),
    ]

submit_slurm_job #

submit_slurm_job(
    command: list[str],
    options: SLURMOptions,
    log_file: Path,
    dependencies: list[str] | None = None,
) -> str

Submit a set of SLURM jobs to the queue

Parameters:

  • command (list[str]) –

    The command to run.

  • options (SLURMOptions) –

    The SLURM options to use.

  • log_file (Path) –

    The file to write the SLURM output to.

  • dependencies (list[str] | None, default: None ) –

    A list of SLURM job IDs to wait for before running this job.

Source code in femto/fe/utils/queue.py 
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
def submit_slurm_job(
    command: list[str],
    options: SLURMOptions,
    log_file: pathlib.Path,
    dependencies: list[str] | None = None,
) -> str:
    """Submit a set of SLURM jobs to the queue

    Args:
        command: The command to run.
        options: The SLURM options to use.
        log_file: The file to write the SLURM output to.
        dependencies: A list of SLURM job IDs to wait for before running this job.
    """

    log_file.parent.mkdir(parents=True, exist_ok=True)

    slurm_args = [f"--output={log_file}", *options.to_flags(), "--parsable"]

    if dependencies is not None and len(dependencies) > 0:
        slurm_args.append("--dependency=afterok:" + ",".join(dependencies))
        slurm_args.append("--kill-on-invalid-dep=yes")

    command_str = shlex.join(
        str(arg) if isinstance(arg, pathlib.Path) else arg for arg in command
    )
    run_args = ["sbatch", *slurm_args, "--wrap", command_str]

    result = subprocess.run(run_args, capture_output=True, text=True, check=True)
    result.check_returncode()

    job_id = result.stdout.strip().strip("\n")

    return job_id

cancel_slurm_jobs #

cancel_slurm_jobs(job_ids: Iterable[str])

Cancel a set of SLURM jobs

Parameters:

  • job_ids (Iterable[str]) –

    The IDs of the jobs to cancel.

Source code in femto/fe/utils/queue.py 
 97
 98
 99
100
101
102
103
104
def cancel_slurm_jobs(job_ids: typing.Iterable[str]):
    """Cancel a set of SLURM jobs

    Args:
        job_ids: The IDs of the jobs to cancel.
    """

    subprocess.run(["scancel", *job_ids])

wait_for_slurm_jobs #

wait_for_slurm_jobs(job_ids: Iterable[str])

Wait for a set of SLURM jobs to finish, or attempt to cancel them if the program fails before they do.

Source code in femto/fe/utils/queue.py 
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
def wait_for_slurm_jobs(job_ids: typing.Iterable[str]):
    """Wait for a set of SLURM jobs to finish, or attempt to cancel them if the
    program fails before they do."""

    original_signal_handlers = {
        signal.SIGINT: signal.getsignal(signal.SIGINT),
        signal.SIGTERM: signal.getsignal(signal.SIGTERM),
    }

    cleanup_func = functools.partial(_cancel_slurm_job_and_exit, job_ids=job_ids)

    for signal_code in original_signal_handlers:
        signal.signal(signal_code, cleanup_func)

    try:
        remaining_job_ids = {*job_ids}

        while True:
            result = subprocess.run(
                ["squeue", "--job", ",".join(remaining_job_ids)],
                capture_output=True,
                text=True,
            )

            if result.returncode != 0:
                _LOGGER.warning(result.stdout)
                _LOGGER.warning(result.stderr)
                signal.raise_signal(signal.SIGTERM)

            finished_jobs = {
                job_id for job_id in remaining_job_ids if job_id not in result.stdout
            }
            remaining_job_ids -= finished_jobs

            if len(remaining_job_ids) == 0:
                break

            time.sleep(5.0)

    finally:
        for signal_code in original_signal_handlers:
            signal.signal(signal_code, original_signal_handlers[signal_code])