Default EarlyStopping.check_on_train_epoch_end=True

CHANGELOG.md

@@ -154,6 +154,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Validation is now always run inside the training epoch scope ([#7357](https://github.com/PyTorchLightning/pytorch-lightning/pull/7357))
 
 
+- `EarlyStopping` now runs at the end of the training epoch by default ([#8286](https://github.com/PyTorchLightning/pytorch-lightning/pull/8286))
+
+
 - Refactored Loops
     * Moved attributes `global_step`, `current_epoch`, `max/min_steps`, `max/min_epochs`, `batch_idx`, and `total_batch_idx` to TrainLoop ([#7437](https://github.com/PyTorchLightning/pytorch-lightning/pull/7437))
     * Refactored result handling in training loop ([#7506](https://github.com/PyTorchLightning/pytorch-lightning/pull/7506))

pytorch_lightning/callbacks/early_stopping.py

@@ -97,7 +97,7 @@ def __init__(
         check_finite: bool = True,
         stopping_threshold: Optional[float] = None,
         divergence_threshold: Optional[float] = None,
-        check_on_train_epoch_end: bool = False,
+        check_on_train_epoch_end: bool = True,
     ):
         super().__init__()
         self.min_delta = min_delta
@@ -149,7 +149,12 @@ def _validate_condition_metric(self, logs):
     def monitor_op(self) -> Callable:
         return self.mode_dict[self.mode]
 
-    def on_save_checkpoint(self, trainer, pl_module, checkpoint: Dict[str, Any]) -> Dict[str, Any]:
+    def on_save_checkpoint(
+        self,
+        trainer: 'pl.Trainer',
+        pl_module: 'pl.LightningModule',
+        checkpoint: Dict[str, Any],
+    ) -> Dict[str, Any]:
         return {
             'wait_count': self.wait_count,
             'stopped_epoch': self.stopped_epoch,
@@ -167,18 +172,17 @@ def _should_skip_check(self, trainer) -> bool:
         from pytorch_lightning.trainer.states import TrainerFn
         return trainer.state.fn != TrainerFn.FITTING or trainer.sanity_checking
 
-    def on_train_epoch_end(self, trainer, pl_module) -> None:
+    def on_train_epoch_end(self, trainer: 'pl.Trainer', pl_module: 'pl.LightningModule') -> None:
         if not self._check_on_train_epoch_end or self._should_skip_check(trainer):
             return
         self._run_early_stopping_check(trainer)
 
     def on_validation_end(self, trainer, pl_module) -> None:
         if self._check_on_train_epoch_end or self._should_skip_check(trainer):
             return
-
         self._run_early_stopping_check(trainer)
 
-    def _run_early_stopping_check(self, trainer) -> None:
+    def _run_early_stopping_check(self, trainer: 'pl.Trainer') -> None:
         """
         Checks whether the early stopping condition is met
         and if so tells the trainer to stop the training.

tests/callbacks/test_early_stopping.py

@@ -69,12 +69,13 @@ def test_resume_early_stopping_from_checkpoint(tmpdir):
     )
     trainer.fit(model, datamodule=dm)
 
+    assert len(early_stop_callback.saved_states) == 4
+
     checkpoint_filepath = checkpoint_callback.kth_best_model_path
     # ensure state is persisted properly
     checkpoint = torch.load(checkpoint_filepath)
     # the checkpoint saves "epoch + 1"
     early_stop_callback_state = early_stop_callback.saved_states[checkpoint["epoch"] - 1]
-    assert 4 == len(early_stop_callback.saved_states)
     assert checkpoint["callbacks"][type(early_stop_callback)] == early_stop_callback_state
 
     # ensure state is reloaded properly (assertion in the callback)