EpochBasedTrainLoop

class mmengine.runner.EpochBasedTrainLoop(runner, dataloader, max_epochs, val_begin=1, val_interval=1, dynamic_intervals=None)

Loop for epoch-based training.

参数

• runner (Runner) – A reference of runner.

• dataloader (Dataloader or dict) – A dataloader object or a dict to build a dataloader.

• max_epochs (int) – Total training epochs.

• val_begin (int) – The epoch that begins validating. Defaults to 1.

• val_interval (int) – Validation interval. Defaults to 1.

• dynamic_intervals (List[Tuple[int, int]], optional) – The first element in the tuple is a milestone and the second element is a interval. The interval is used after the corresponding milestone. Defaults to None.

返回类型

None

property epoch

Current epoch.

Type

int

property iter

Current iteration.

Type

int

property max_epochs

Total epochs to train model.

Type

int

property max_iters

Total iterations to train model.

Type

int

run()

Launch training.

返回类型

torch.nn.modules.module.Module

run_epoch()

Iterate one epoch.

返回类型

None

run_iter(idx, data_batch)

Iterate one min-batch.

参数

data_batch (Sequence[dict]) – Batch of data from dataloader.

返回类型

None