core._cli_hydra

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree.

Attributes

ALLOWED_TOP_LEVEL_KEYS
LOG_DIR_NAME
CHECKPOINT_DIR_NAME
RESULTS_DIR
CONFIG_FILE_NAME
PREEMPTION_STATE_DIR_NAME

Classes

SchedulerType	str(object='') -> str
DeviceType	str(object='') -> str
RunType	str(object='') -> str
SlurmConfig
SchedulerConfig
SlurmEnv
Metadata
JobConfig
Submitit	Derived callable classes are requeued after timeout with their current

Functions

_set_seeds(→ None)
_set_deterministic_mode(→ None)
_get_slurm_env(→ SlurmEnv)
remove_runner_state_from_submission(→ None)
map_job_config_to_dist_config(→ dict)
get_canonical_config(→ omegaconf.DictConfig)
get_hydra_config_from_yaml(→ omegaconf.DictConfig)
_runner_wrapper(config[, run_type])
main([args, override_args])

Module Contents

core._cli_hydra.ALLOWED_TOP_LEVEL_KEYS

core._cli_hydra.LOG_DIR_NAME = 'logs'

core._cli_hydra.CHECKPOINT_DIR_NAME = 'checkpoints'

core._cli_hydra.RESULTS_DIR = 'results'

core._cli_hydra.CONFIG_FILE_NAME = 'canonical_config.yaml'

core._cli_hydra.PREEMPTION_STATE_DIR_NAME = 'preemption_state'

class core._cli_hydra.SchedulerType

Bases: str, enum.Enum

str(object=’’) -> str str(bytes_or_buffer[, encoding[, errors]]) -> str

Create a new string object from the given object. If encoding or errors is specified, then the object must expose a data buffer that will be decoded using the given encoding and error handler. Otherwise, returns the result of object.__str__() (if defined) or repr(object). encoding defaults to sys.getdefaultencoding(). errors defaults to ‘strict’.

LOCAL = 'local'

SLURM = 'slurm'

class core._cli_hydra.DeviceType

Bases: str, enum.Enum

str(object=’’) -> str str(bytes_or_buffer[, encoding[, errors]]) -> str

Create a new string object from the given object. If encoding or errors is specified, then the object must expose a data buffer that will be decoded using the given encoding and error handler. Otherwise, returns the result of object.__str__() (if defined) or repr(object). encoding defaults to sys.getdefaultencoding(). errors defaults to ‘strict’.

CPU = 'cpu'

CUDA = 'cuda'

class core._cli_hydra.RunType

Bases: str, enum.Enum

str(object=’’) -> str str(bytes_or_buffer[, encoding[, errors]]) -> str

Create a new string object from the given object. If encoding or errors is specified, then the object must expose a data buffer that will be decoded using the given encoding and error handler. Otherwise, returns the result of object.__str__() (if defined) or repr(object). encoding defaults to sys.getdefaultencoding(). errors defaults to ‘strict’.

RUN = 'run'

REDUCE = 'reduce'

class core._cli_hydra.SlurmConfig

mem_gb: int = 80

timeout_hr: int = 168

cpus_per_task: int = 8

partition: str | None = None

qos: str | None = None

account: str | None = None

class core._cli_hydra.SchedulerConfig

mode: SchedulerType

ranks_per_node: int = 1

num_nodes: int = 1

num_array_jobs: int = 1

slurm: SlurmConfig

class core._cli_hydra.SlurmEnv

job_id: str | None = None

raw_job_id: str | None = None

array_job_id: str | None = None

array_task_id: str | None = None

restart_count: str | None = None

class core._cli_hydra.Metadata

commit: str

log_dir: str

checkpoint_dir: str

results_dir: str

config_path: str

preemption_checkpoint_dir: str

cluster_name: str

array_job_num: int = 0

slurm_env: SlurmEnv

class core._cli_hydra.JobConfig

run_name: str

timestamp_id: str

run_dir: str

device_type: DeviceType

debug: bool = False

scheduler: SchedulerConfig

logger: dict | None = None

seed: int = 0

deterministic: bool = False

runner_state_path: str | None = None

metadata: Metadata | None = None

graph_parallel_group_size: int | None = None

__post_init__() → None

core._cli_hydra._set_seeds(seed: int) → None

core._cli_hydra._set_deterministic_mode() → None

core._cli_hydra._get_slurm_env() → SlurmEnv

core._cli_hydra.remove_runner_state_from_submission(log_folder: str, job_id: str) → None

class core._cli_hydra.Submitit

Bases: submitit.helpers.Checkpointable

Derived callable classes are requeued after timeout with their current state dumped at checkpoint.

__call__ method must be implemented to make your class a callable.

Note

The following implementation of the checkpoint method resubmits the full current state of the callable (self) with the initial argument. You may want to replace the method to curate the state (dump a neural network to a standard format and remove it from the state so that not to pickle it) and change/remove the initial parameters.

config = None

runner = None

__call__(dict_config: omegaconf.DictConfig, run_type: RunType = RunType.RUN) → None

_init_logger() → None

checkpoint(*args, **kwargs) → submitit.helpers.DelayedSubmission

Resubmits the same callable with the same arguments

core._cli_hydra.map_job_config_to_dist_config(job_cfg: JobConfig) → dict

core._cli_hydra.get_canonical_config(config: omegaconf.DictConfig) → omegaconf.DictConfig

core._cli_hydra.get_hydra_config_from_yaml(config_yml: str, overrides_args: list[str]) → omegaconf.DictConfig

core._cli_hydra._runner_wrapper(config: omegaconf.DictConfig, run_type: RunType = RunType.RUN)

core._cli_hydra.main(args: argparse.Namespace | None = None, override_args: list[str] | None = None)