Close SummaryWriter in TensorBoardLogger on finalize (#5696)

Not entirely sure this is the "right" solution to this problem, but currently when model fitting is finished the `TensorBoardLogger` attribute `_experiment` (a `SummaryWriter`) is left with an open file handle. This causes issues in particular on Windows systems (and probably others), and also makes the files un-syncable on cloud-synced devices like OneDrive. This PR adds a `close()` to `finalize` to make sure this handle is closed upon fit completion.

Co-authored-by: Adrian Wälchli <[email protected]>

Author: 2 people

CHANGELOG.md

@@ -188,6 +188,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Fixed
 
+- Fixed `TensorBoardLogger` not closing `SummaryWriter` on `finalize` ([#5696](https://github.com/PyTorchLightning/pytorch-lightning/pull/5696))
+
+
 - Fixed filtering of pytorch  "unsqueeze" warning when using DP ([#5622](https://github.com/PyTorchLightning/pytorch-lightning/pull/5622))
 
 

pytorch_lightning/loggers/tensorboard.py

@@ -237,6 +237,7 @@ def save(self) -> None:
     @rank_zero_only
     def finalize(self, status: str) -> None:
         self.experiment.flush()
+        self.experiment.close()
         self.save()
 
     @property

tests/loggers/test_tensorboard.py

@@ -265,3 +265,12 @@ def configure_optimizers(self):
 
     mock_count_steps = [m[2]["step"] for m in mock_log_metrics.mock_calls if "count_step" in m[2]["metrics"]]
     assert model._indexes == mock_count_steps
+
+
+@mock.patch('pytorch_lightning.loggers.tensorboard.SummaryWriter')
+def test_tensorboard_finalize(summary_writer, tmpdir):
+    """ Test that the SummaryWriter closes in finalize. """
+    logger = TensorBoardLogger(save_dir=tmpdir)
+    logger.finalize("any")
+    summary_writer().flush.assert_called()
+    summary_writer().close.assert_called()