Forward verbose status to tuning optimization (#214)

* forward verbose status to tuning optimization

* Update exp_manager.py

* Update changelog and add comment

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

Author: jkterry1

CHANGELOG.md

@@ -7,6 +7,7 @@
 - Support experiment tracking via Weights and Biases (@vwxyzjn)
 
 ### New Features
+- Verbose mode for each trial (when doing hyperparam optimization) can now be activated using the debug mode (verbose == 2)
 
 ### Bug fixes
 - Policies saved during during optimization with distributed Optuna load on new systems (@jkterry)
@@ -16,7 +17,6 @@
 
 ### Other
 
-
 ## Release 1.4.0 (2022-01-19)
 
 ### Breaking Changes

utils/exp_manager.py

@@ -628,12 +628,20 @@ def objective(self, trial: optuna.Trial) -> float:
         n_envs = 1 if self.algo == "ars" else self.n_envs
         env = self.create_envs(n_envs, no_log=True)
 
+        # By default, do not activate verbose output to keep
+        # stdout clean with only the trials results
+        trial_verbosity = 0
+        # Activate verbose mode for the trial in debug mode
+        # See PR #214
+        if self.verbose >= 2:
+            trial_verbosity = self.verbose
+
         model = ALGOS[self.algo](
             env=env,
             tensorboard_log=None,
             # We do not seed the trial
             seed=None,
-            verbose=0,
+            verbose=trial_verbosity,
             **kwargs,
         )