Skip to navigation

nemo_gym.orchestration.executors.slurm

View as MarkdownOpen in Claude

The Slurm executor for Pyxis-enabled clusters: SlurmExecutor.

Its parts, each in its own module:

  • executor: render a run’s job scripts, submit them with sbatch and record the submission;
  • chain: recovery of a run’s auto-resume chain (recovery_evidence, resubmit), mixed into the executor;
  • judge_job: judge-only jobs that grade a run’s saved answers again (reverify_judge_failures), mixed in too;
  • reader: the remote Store reader both of them read a run through;
  • scheduler: the Slurm commands they share;
  • script: render a benchmark’s sbatch script, and resume_script: the auto-resume prologue and epilogue it includes for a resumable job;
  • lustre_md_stats: read the per-phase Lustre metadata counters a job records with lustre_md_stats: true.

Only Slurm implements resubmit today, so these parts are Slurm code with no scheduler-agnostic interface between them; the executor capabilities give resubmit its seam when a second scheduler needs one.

Submodules

Package Contents

Classes

NameDescription
SlurmExecutorSlurm executor for Pyxis-enabled clusters (https://github.com/NVIDIA/pyxis).

API

class nemo_gym.orchestration.executors.slurm.SlurmExecutor()

Bases: ChainRecovery, JudgeJobs, BaseExecutor, Resubmittable, Terminable, LogReadable, Inspectable

Slurm executor for Pyxis-enabled clusters (https://github.com/NVIDIA/pyxis).

Every service and the driver are launched via srun --container-image so they run inside the container specified in their config. Health checks run as plain bash inside the sbatch script (no container needed — they just poll HTTP).

name
= 'slurm'
nemo_gym.orchestration.executors.slurm.SlurmExecutor._build_record(
cluster: str,
gym_job_id: str,
now: datetime.datetime,
remote_run_dir: pathlib.Path,
benchmark_names: list[str],
output: str,
nemo_gym.orchestration.executors.slurm.SlurmExecutor._dry_run(
remote_run_dir: pathlib.Path
) -> None
nemo_gym.orchestration.executors.slurm.SlurmExecutor._saved_config(
run_dir: pathlib.Path
staticmethod

The resolved config persist() saved for this run, so a resubmit repeats what submit did.

nemo_gym.orchestration.executors.slurm.SlurmExecutor._stage(
remote_run_dir: pathlib.Path,
staging: pathlib.Path
) -> pathlib.Path
nemo_gym.orchestration.executors.slurm.SlurmExecutor.connection_key(
) -> str
nemo_gym.orchestration.executors.slurm.SlurmExecutor.export_tracking(
benchmark: str,
) -> dict[str, typing.Any]

Run gym eval export mlflow on the run’s saved artifacts from the login host, and report its result.

The export runs where the remote Store reader runs, with the run’s saved pinned Gym install (and its MLflow extra) in a temporary checkout: it only reads the saved artifacts and talks to the tracking server, so it needs no allocation. It takes the benchmark’s saved mlflow_* and upload_rollouts run settings, and its credentials as the driver did: the submission’s forwarded environment and the driver’s env. Gym’s export takes the run lock, prepares the health metrics and writes the receipt bound to the artifact digests, and skips a run that already has one. The resubmit lock is held and no job of the chain may be live meanwhile.

nemo_gym.orchestration.executors.slurm.SlurmExecutor.gym_pin(
) -> str | None

The full Gym commit the run’s saved config pins its driver to; None when there is none or it is unreadable.

nemo_gym.orchestration.executors.slurm.SlurmExecutor.installed_commit() -> str | None

The Gym commit this launcher runs; the same value the judge-only job and a repin are checked against.

nemo_gym.orchestration.executors.slurm.SlurmExecutor.query_states(

One script on one connection: a single squeue, a single sacct and each job directory’s state files.

nemo_gym.orchestration.executors.slurm.SlurmExecutor.read_run_files(
paths: list[pathlib.Path]
) -> dict[pathlib.Path, str | None]
nemo_gym.orchestration.executors.slurm.SlurmExecutor.run(
dry_run: bool = False
nemo_gym.orchestration.executors.slurm.SlurmExecutor.stream_logs(
tail: int | None,
follow: bool
) -> collections.abc.Iterator[str]

The lines of source, read in one call; the location is relative to the job directory.

nemo_gym.orchestration.executors.slurm.SlurmExecutor.terminate(
) -> None

Cancel everything of job that is queued or running, the pending jobs first, in one script.