phlower.settings.PhlowerTrainerSetting

class phlower.settings.PhlowerTrainerSetting(*, loss_setting=<factory>, optimizer_setting=<factory>, scheduler_settings=<factory>, handler_settings=<factory>, n_epoch=10, random_seed=0, batch_size=1, num_workers=0, device='cpu', evaluation_for_training=True, log_every_n_epoch=1, initializer_setting=<factory>, lazy_load=True, time_series_sliding=<factory>, parallel_setting=<factory>, non_blocking=False, pin_memory=False)

Bases: BaseModel

Methods

get_device([rank])
get_early_stopping_patience()

Attributes

model_computed_fields
model_config	Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].
model_extra	Get extra fields set during validation.
model_fields
model_fields_set	Returns the set of fields that have been explicitly set on this model instance.
loss_setting	setting for loss function
optimizer_setting	setting for optimizer
scheduler_settings	setting for schedulers
handler_settings	setting for handlers
n_epoch	the number of epochs.
random_seed	random seed.
batch_size	batch size.
num_workers	the number of cores.
device	device name.
evaluation_for_training	If True, evaluation for training dataset is performed
log_every_n_epoch	dump log items every nth epoch
initializer_setting	setting for trainer initializer
lazy_load	If True, data is loaded lazily.
time_series_sliding	Setting for sliding window for time series data.
parallel_setting	Setting for parallel processing
non_blocking	If True, non_blocking transfer is used when data is transferred to device.
pin_memory	If True, pin_memory is used in DataLoader.

Parameters:

• loss_setting (LossSetting)

• optimizer_setting (OptimizerSetting)

• scheduler_settings (list[SchedulerSetting])

• handler_settings (list[Annotated[Annotated[EarlyStoppingSetting, Tag(tag=EarlyStopping)] | Annotated[UserDefinedHandlerSetting, Tag(tag=UserCustom)], Discriminator(discriminator=~phlower.settings._handler_settings._custom_handler_discriminator, custom_error_type=invalid_union_member, custom_error_message=Invalid union member, custom_error_context={'discriminator': 'handler_checkk'})]])

• n_epoch (int)

• random_seed (int)

• batch_size (int)

• num_workers (int)

• device (str)

• evaluation_for_training (bool)

• log_every_n_epoch (int)

• initializer_setting (TrainerInitializerSetting)

• lazy_load (bool)

• time_series_sliding (TimeSeriesSlidingSetting)

• parallel_setting (ParallelSetting)

• non_blocking (bool)

• pin_memory (bool)

batch_size: int

batch size. Defaults to 1

device: str

device name. Defaults to cpu. When auto is set, device is automatically set to gpu when available, otherwise cpu.

evaluation_for_training: bool

If True, evaluation for training dataset is performed

handler_settings: list[HandlerSettingType]

setting for handlers

initializer_setting: TrainerInitializerSetting

setting for trainer initializer

lazy_load: bool

If True, data is loaded lazily. If False, all data is loaded at once. Defaults to True.

log_every_n_epoch: int

dump log items every nth epoch

loss_setting: LossSetting

setting for loss function

model_config: ClassVar[ConfigDict] = {'extra': 'forbid', 'frozen': True}

Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

n_epoch: int

the number of epochs. Defaults to 10.

non_blocking: bool

If True, non_blocking transfer is used when data is transferred to device. Defaults to False.

num_workers: int

the number of cores. Defaults to 0.

optimizer_setting: OptimizerSetting

setting for optimizer

parallel_setting: ParallelSetting

Setting for parallel processing

pin_memory: bool

If True, pin_memory is used in DataLoader. Defaults to False.

random_seed: int

random seed. Defaults to 0

scheduler_settings: list[SchedulerSetting]

setting for schedulers

time_series_sliding: TimeSeriesSlidingSetting

Setting for sliding window for time series data. Defaults to inactive.