added warning when changing monitor and using results obj (Lightning-…

…AI#3014)

* added warning when changing monitor and using results obj

* added warning when changing monitor and using results obj

* added warning when changing monitor and using results obj

pytorch_lightning/callbacks/early_stopping.py

@@ -14,6 +14,7 @@
 from pytorch_lightning import _logger as log
 from pytorch_lightning.callbacks.base import Callback
 from pytorch_lightning.utilities import rank_zero_warn
+import os
 
 torch_inf = torch.tensor(np.Inf)
 
@@ -72,6 +73,7 @@ def __init__(self, monitor: str = 'val_loss', min_delta: float = 0.0, patience:
         self.wait_count = 0
         self.stopped_epoch = 0
         self.mode = mode
+        self.warned_result_obj = False
 
         if mode not in self.mode_dict:
             if self.verbose > 0:
@@ -154,12 +156,26 @@ def on_train_epoch_end(self, trainer, pl_module):
         if should_check_early_stop:
             self._run_early_stopping_check(trainer, pl_module)
 
+    def __warn_deprecated_monitor_key(self):
+        using_result_obj = os.environ.get('PL_USING_RESULT_OBJ', None)
+        invalid_key = self.monitor not in ['val_loss', 'early_stop_on', 'val_early_step_on']
+        if using_result_obj and not self.warned_result_obj and invalid_key:
+            self.warned_result_obj = True
+            m = f"""
+                    When using EvalResult(early_stop_on=X) or TrainResult(early_stop_on=X) the
+                    'monitor' key of EarlyStopping has no effect.
+                    Remove EarlyStopping(monitor='{self.monitor}) to fix')
+                """
+            rank_zero_warn(m)
+
     def _run_early_stopping_check(self, trainer, pl_module):
         logs = trainer.callback_metrics
 
         if not self._validate_condition_metric(logs):
             return  # short circuit if metric not present
 
+        self.__warn_deprecated_monitor_key()
+
         current = logs.get(self.monitor)
 
         # when in dev debugging

pytorch_lightning/callbacks/model_checkpoint.py

@@ -138,6 +138,7 @@ def __init__(self, filepath: Optional[str] = None, monitor: str = 'val_loss', ve
         self.best_model_score = 0
         self.best_model_path = ''
         self.save_function = None
+        self.warned_result_obj = False
 
         torch_inf = torch.tensor(np.Inf)
         mode_dict = {
@@ -297,12 +298,27 @@ def on_train_start(self, trainer, pl_module):
         if not gfile.exists(self.dirpath):
             makedirs(self.dirpath)
 
+    def __warn_deprecated_monitor_key(self):
+        using_result_obj = os.environ.get('PL_USING_RESULT_OBJ', None)
+        invalid_key = self.monitor not in ['val_loss', 'checkpoint_on']
+        if using_result_obj and not self.warned_result_obj and invalid_key:
+            self.warned_result_obj = True
+            m = f"""
+                    When using EvalResult(early_stop_on=X) or TrainResult(early_stop_on=X) the
+                    'monitor' key of ModelCheckpoint has no effect.
+                    Remove ModelCheckpoint(monitor='{self.monitor}) to fix')
+                """
+            rank_zero_warn(m)
+
     @rank_zero_only
     def on_validation_end(self, trainer, pl_module):
         # only run on main process
         if trainer.global_rank != 0:
             return
 
+        # TODO: remove when dict results are deprecated
+        self.__warn_deprecated_monitor_key()
+
         metrics = trainer.callback_metrics
         epoch = trainer.current_epoch
 

pytorch_lightning/core/step_result.py

@@ -4,6 +4,7 @@
 
 import torch
 from torch import Tensor
+import os
 
 from pytorch_lightning.metrics.converters import _sync_ddp_if_available
 
@@ -20,6 +21,9 @@ def __init__(
 
         super().__init__()
 
+        # temporary until dict results are deprecated
+        os.environ['PL_USING_RESULT_OBJ'] = '1'
+
         if early_stop_on is not None:
             self.early_stop_on = early_stop_on
         if checkpoint_on is not None and checkpoint_on:

tests/trainer/test_trainer_steps_result_return.py

@@ -7,6 +7,7 @@
 import torch
 
 from pytorch_lightning import Trainer
+from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint
 from pytorch_lightning.core.step_result import TrainResult
 from tests.base import EvalModelTemplate
 from tests.base.deterministic_model import DeterministicModel
@@ -543,3 +544,43 @@ def test_result_map(tmpdir):
     assert 'x2' not in result
     assert 'y1' in result
     assert 'y2' in result
+
+
+def test_result_monitor_warnings(tmpdir):
+    """
+    Tests that we warn when the monitor key is changed and we use Results obj
+    """
+    model = EvalModelTemplate()
+    model.test_step = None
+    model.training_step = model.training_step_result_obj
+    model.training_step_end = None
+    model.training_epoch_end = None
+    model.validation_step = model.validation_step_result_obj
+    model.validation_step_end = None
+    model.validation_epoch_end = None
+    model.test_dataloader = None
+
+    trainer = Trainer(
+        default_root_dir=tmpdir,
+        max_epochs=2,
+        early_stop_callback=True,
+        row_log_interval=2,
+        limit_train_batches=2,
+        weights_summary=None,
+        checkpoint_callback=ModelCheckpoint(monitor='not_val_loss')
+    )
+
+    with pytest.warns(UserWarning, match='key of ModelCheckpoint has no effect'):
+        trainer.fit(model)
+
+    trainer = Trainer(
+        default_root_dir=tmpdir,
+        max_epochs=2,
+        row_log_interval=2,
+        limit_train_batches=2,
+        weights_summary=None,
+        early_stop_callback=EarlyStopping(monitor='not_val_loss')
+    )
+
+    with pytest.warns(UserWarning, match='key of EarlyStopping has no effec'):
+        trainer.fit(model)