Early stopping when validation is disabled

CHANGELOG.md

@@ -43,6 +43,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed all warnings and errors in the docs build process ([#1191](https://github.com/PyTorchLightning/pytorch-lightning/pull/1191))
 - Fixed an issue where `val_percent_check=0` would not disable validation ([#1251](https://github.com/PyTorchLightning/pytorch-lightning/pull/1251))
 - Fixed average of incomplete `TensorRunningMean` ([#1309](https://github.com/PyTorchLightning/pytorch-lightning/pull/1309))
+- Fixed an issue with early stopping that would prevent it from monitoring training metrics when validation is disabled / not implemented ([#1235](https://github.com/PyTorchLightning/pytorch-lightning/pull/1235)). 
 
 ## [0.7.1] - 2020-03-07
 

docs/source/early_stopping.rst

@@ -4,8 +4,8 @@ Early stopping
 Default behavior
 ----------------
 By default early stopping will be enabled if `'val_loss'`
-is found in `validation_epoch_end()` return dict. Otherwise
-training will proceed with early stopping disabled.
+is found in :meth:`~pytorch_lightning.core.lightning.LightningModule.validation_epoch_end`'s
+return dict. Otherwise training will proceed with early stopping disabled.
 
 Enable Early Stopping
 ---------------------
@@ -30,9 +30,18 @@ There are two ways to enable early stopping.
     )
     trainer = Trainer(early_stop_callback=early_stop_callback)
 
+In any case, the callback will fall back to the training metrics (returned in
+:meth:`~pytorch_lightning.core.lightning.LightningModule.training_step`,
+:meth:`~pytorch_lightning.core.lightning.LightningModule.training_step_end`)
+looking for a key to monitor if validation is disabled or
+:meth:`~pytorch_lightning.core.lightning.LightningModule.validation_epoch_end`
+is not defined.
+
+
 Disable Early Stopping
 ----------------------
-To disable early stopping pass ``False`` to the `early_stop_callback`.
+To disable early stopping pass ``False`` to the
+:paramref:`~pytorch_lightning.trainer.trainer.Trainer.early_stop_callback`.
 Note that ``None`` will not disable early stopping but will lead to the
 default behaviour.
 

pytorch_lightning/trainer/training_loop.py

@@ -367,8 +367,8 @@ def train(self):
                 met_min_steps = self.global_step >= self.min_steps if self.min_steps else True
 
                 # TODO wrap this logic into the callback
-                if self.enable_early_stop and not self.disable_validation and is_val_epoch:
-                    if ((met_min_epochs and met_min_steps) or self.fast_dev_run):
+                if self.enable_early_stop:
+                    if (met_min_epochs and met_min_steps) or self.fast_dev_run:
                         should_stop = self.early_stop_callback.on_epoch_end(self, self.get_model())
                         # stop training
                         stop = should_stop and met_min_epochs

tests/trainer/test_callbacks.py

@@ -1,11 +1,12 @@
 import tests.base.utils as tutils
 from pytorch_lightning import Callback
 from pytorch_lightning import Trainer, LightningModule
+from pytorch_lightning.callbacks import EarlyStopping
 from tests.base import (
-    TestModelBase,
     LightTrainDataloader,
+    LightTestMixin,
     LightValidationMixin,
-    LightTestMixin
+    TestModelBase
 )
 
 
@@ -150,3 +151,33 @@ def on_test_end(self, trainer, pl_module):
 
     assert test_callback.on_test_start_called
     assert test_callback.on_test_end_called
+
+
+def test_early_stopping_without_val_step(tmpdir):
+    """Test that early stopping callback falls back to training metrics when no validation defined."""
+    tutils.reset_seed()
+
+    class ModelWithoutValStep(LightTrainDataloader, TestModelBase):
+
+        def training_step(self, *args, **kwargs):
+            output = super().training_step(*args, **kwargs)
+            loss = output['loss']  # could be anything else
+            output.update({'my_train_metric': loss})
+            return output
+
+    hparams = tutils.get_default_hparams()
+    model = ModelWithoutValStep(hparams)
+
+    stopping = EarlyStopping(monitor='my_train_metric', min_delta=0.1)
+    trainer_options = dict(
+        default_save_path=tmpdir,
+        early_stop_callback=stopping,
+        overfit_pct=0.20,
+        max_epochs=10,
+    )
+
+    trainer = Trainer(**trainer_options)
+    result = trainer.fit(model)
+
+    assert result == 1, 'training failed to complete'
+    assert trainer.current_epoch < trainer.max_epochs - 1