Fix ProgressBarCallback under-reporting (#1260)

* Updated tqdm progress bar constructor to account for the effects of train_freq/n_steps/num_envs on total_timesteps. Ensure progress bar is "flushed" on training end.

* Added description of PR #1260. Fixed formatting typo

* Partial revert

Co-authored-by: dominicgkerr <[email protected]>
Co-authored-by: Antonin RAFFIN <[email protected]>
Co-authored-by: Antonin Raffin <[email protected]>

Author: 3 people

docs/misc/changelog.rst

@@ -53,6 +53,7 @@ New Features:
 
 Bug Fixes:
 ^^^^^^^^^^
+- Updated ``ProgressBarCallback`` to more accurately display the proportion of remaining timesteps (@dominicgkerr)
 - Fixed return type of ``evaluate_actions`` in ``ActorCritcPolicy`` to reflect that entropy is an optional tensor (@Rocamonde)
 - Fixed type annotation of ``policy`` in ``BaseAlgorithm`` and ``OffPolicyAlgorithm``
 - Allowed model trained with Python 3.7 to be loaded with Python 3.8+ without the ``custom_objects`` workaround
@@ -180,7 +181,7 @@ Others:
 ^^^^^^^
 - Fixed ``DictReplayBuffer.next_observations`` typing (@qgallouedec)
 - Added support for ``device="auto"`` in buffers and made it default (@qgallouedec)
-- Updated ``ResultsWriter` (used internally by ``Monitor`` wrapper) to automatically create missing directories when ``filename`` is a path (@dominicgkerr)
+- Updated ``ResultsWriter`` (used internally by ``Monitor`` wrapper) to automatically create missing directories when ``filename`` is a path (@dominicgkerr)
 
 Documentation:
 ^^^^^^^^^^^^^^

stable_baselines3/common/callbacks.py

@@ -685,5 +685,6 @@ def _on_step(self) -> bool:
         return True
 
     def _on_training_end(self) -> None:
-        # Close progress bar
+        # Flush and close progress bar
+        self.pbar.refresh()
         self.pbar.close()