nemo_gym.orchestration.executors.kubernetes

Kubernetes executor: one kubectl apply per benchmark, no SSH/rsync.

Assumes the kubeconfig is already set up (e.g. tsh kube login <cluster>) the same way the Slurm executor assumes SSH keys/agent are already set up — this executor does not drive cluster auth.

Module Contents

Classes

Name	Description
`KubernetesExecutor`	Kubernetes executor. One Job per benchmark; see `kubernetes_script.py` for the pod shape.

Functions

Name	Description
`_kubectl`	-
`_validate_benchmark_names`	-

Data

_KUBECTL_TIMEOUT_SECONDS

_VALID_BENCHMARK_NAME

API

class nemo_gym.orchestration.executors.kubernetes.KubernetesExecutor()

Bases: BaseExecutor

Kubernetes executor. One Job per benchmark; see kubernetes_script.py for the pod shape.

name

= 'kubernetes'

nemo_gym.orchestration.executors.kubernetes.KubernetesExecutor._build_manifests(
    config: nemo_gym.orchestration.api.SubmitConfig,
    compute: nemo_gym.orchestration.api.KubernetesComputeConfig,
    gym_job_id: str,
    now: datetime.datetime,
    base_run_dir: pathlib.Path,
    benchmark_names: list[str]
) -> list[tuple[str, dict]]

nemo_gym.orchestration.executors.kubernetes.KubernetesExecutor._dry_run(
    manifests: list[tuple[str, dict]],
    gym_job_id: str
) -> None

nemo_gym.orchestration.executors.kubernetes.KubernetesExecutor.resubmit(
    record: nemo_gym.orchestration.jobs.SubmissionRecord,
    benchmark: str,
    keep_budget: bool = False,
    accounted_walltime_s: float | None = None,
    accounting_snapshot: collections.abc.Mapping[str, str | None] | None = None,
    accounting_reconciled: bool = False,
    repin: str | None = None,
    repin_reason: str | None = None,
    extend_retries: int = 0,
    extend_attempts: int = 0,
    max_walltime: str | None = None,
    walltime_reason: str | None = None,
    fresh_restart: bool = False,
    rerun_masked: collections.abc.Mapping[str, typing.Any] | None = None
) -> nemo_gym.orchestration.jobs.BenchmarkJob

nemo_gym.orchestration.executors.kubernetes.KubernetesExecutor.run(
    config: nemo_gym.orchestration.api.SubmitConfig,
    dry_run: bool = False
) -> nemo_gym.orchestration.jobs.SubmissionRecord | None

nemo_gym.orchestration.executors.kubernetes._kubectl(
    compute: nemo_gym.orchestration.api.KubernetesComputeConfig,
    args: str = (),
    input: str | None = None
) -> subprocess.CompletedProcess

nemo_gym.orchestration.executors.kubernetes._validate_benchmark_names(
    benchmarks: list[str]
) -> None

nemo_gym.orchestration.executors.kubernetes._KUBECTL_TIMEOUT_SECONDS = 60

nemo_gym.orchestration.executors.kubernetes._VALID_BENCHMARK_NAME = re.compile('^[a-z0-9][a-z0-9._-]*$')