nemo_automodel.components.launcher.skypilot.launcher

View as Markdown

Module Contents

Classes

NameDescription
SkyPilotLauncherLaunch a recipe job on a cloud VM via SkyPilot.

Functions

NameDescription
_parse_gpus_per_nodeExtract GPU count from an accelerator string like 'A100:8'.
_recipe_module_path-

Data

logger

API

class nemo_automodel.components.launcher.skypilot.launcher.SkyPilotLauncher()

Bases: Launcher

Launch a recipe job on a cloud VM via SkyPilot.

nemo_automodel.components.launcher.skypilot.launcher.SkyPilotLauncher._build_command(
recipe_target: str,
job_conf_path: str,
gpus_per_node: int,
num_nodes: int,
extra_args: typing.Optional[typing.List[str]] = None
) -> str
nemo_automodel.components.launcher.skypilot.launcher.SkyPilotLauncher.launch(
config: typing.Dict[str, typing.Any],
config_path: pathlib.Path,
recipe_target: str,
launcher_config: typing.Dict[str, typing.Any],
extra_args: typing.Optional[typing.List[str]] = None
) -> int
nemo_automodel.components.launcher.skypilot.launcher._parse_gpus_per_node(
accelerators: str
) -> int

Extract GPU count from an accelerator string like 'A100:8'.

Returns 1 when the string cannot be parsed.

nemo_automodel.components.launcher.skypilot.launcher._recipe_module_path(
recipe_target: str,
repo_root: str
) -> str
nemo_automodel.components.launcher.skypilot.launcher.logger = logging.getLogger(__name__)