EvaluationEpochLoop

class pytorch_lightning.loops.epoch.EvaluationEpochLoop

Bases: abc.ABC, Generic[pytorch_lightning.loops.base.T]

This is the loop performing the evaluation.

It mainly loops over the given dataloader and runs the validation or test step (depending on the trainer’s current state).

advance(data_fetcher, dataloader_idx, dl_max_batches, num_dataloaders)

Calls the evaluation step with the corresponding hooks and updates the logger connector.

Parameters

• data_fetcher (AbstractDataFetcher) – iterator over the dataloader

• dataloader_idx (int) – index of the current dataloader

• dl_max_batches (int) – maximum number of batches the dataloader can produce

• num_dataloaders (int) – the total number of dataloaders

Raises

StopIteration – If the current batch is None

Return type

None

connect(**kwargs)

Optionally connect one or multiple loops to this one.

Linked loops should form a tree.

Return type

None

on_load_checkpoint(state_dict)

Called when loading a model checkpoint, use to reload loop state.

Return type

None

on_run_end()

Returns the outputs of the whole run.

Return type

List[Union[Tensor, Dict[str, Any]]]

on_run_start(data_fetcher, dataloader_idx, dl_max_batches, num_dataloaders)

Adds the passed arguments to the loop’s state if necessary.

Parameters

• data_fetcher (AbstractDataFetcher) – the current data_fetcher wrapping the dataloader

• dataloader_idx (int) – index of the current dataloader

• dl_max_batches (int) – maximum number of batches the dataloader can produce

• num_dataloaders (int) – the total number of dataloaders

Return type

None

on_save_checkpoint()

Called when saving a model checkpoint, use to persist loop state.

Return type

Dict

Returns

The current loop state.

reset()

Resets the loop’s internal state.

Return type

None

teardown()

Use to release memory etc.

Return type

None

property done: bool

Returns True if the current iteration count reaches the number of dataloader batches.