fix info message when max training time reached (#7780)

* call time_elapsed

* elapsed formatting

* format

* update test

* changelog

CHANGELOG.md

@@ -19,6 +19,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed print errors in `ProgressBar` when `trainer.fit` is not called ([#7674](https://github.com/PyTorchLightning/pytorch-lightning/pull/7674))
 - Fixed global step update when the epoch is skipped ([#7677](https://github.com/PyTorchLightning/pytorch-lightning/pull/7677))
 - Fixed training loop total batch counter when accumulate grad batches was enabled ([#7692](https://github.com/PyTorchLightning/pytorch-lightning/pull/7692))
+- Fixed formatting of info message when max training time reached ([#7780](https://github.com/PyTorchLightning/pytorch-lightning/pull/7780))
 
 ## [1.3.2] - 2021-05-18
 

pytorch_lightning/callbacks/timer.py

@@ -170,4 +170,5 @@ def _check_time_remaining(self, trainer: 'pl.Trainer') -> None:
         should_stop = trainer.accelerator.broadcast(should_stop)
         trainer.should_stop = trainer.should_stop or should_stop
         if should_stop and self._verbose:
-            rank_zero_info(f"Time limit reached. Elapsed time is {self.time_elapsed}. Signaling Trainer to stop.")
+            elapsed = timedelta(seconds=int(self.time_elapsed(RunningStage.TRAINING)))
+            rank_zero_info(f"Time limit reached. Elapsed time is {elapsed}. Signaling Trainer to stop.")

tests/callbacks/test_timer.py

@@ -95,7 +95,7 @@ def test_timer_time_remaining(time_mock):
     assert round(timer.time_elapsed()) == 3
 
 
-def test_timer_stops_training(tmpdir):
+def test_timer_stops_training(tmpdir, caplog):
     """ Test that the timer stops training before reaching max_epochs """
     model = BoringModel()
     duration = timedelta(milliseconds=100)
@@ -106,9 +106,12 @@ def test_timer_stops_training(tmpdir):
         max_epochs=1000,
         callbacks=[timer],
     )
-    trainer.fit(model)
+    with caplog.at_level(logging.INFO):
+        trainer.fit(model)
     assert trainer.global_step > 1
     assert trainer.current_epoch < 999
+    assert "Time limit reached." in caplog.text
+    assert "Signaling Trainer to stop." in caplog.text
 
 
 @pytest.mark.parametrize("interval", ["step", "epoch"])