Log success rate for PPO variants

sb3_contrib/ppo_mask/ppo_mask.py

@@ -1,5 +1,3 @@
-import sys
-import time
 from typing import Any, ClassVar, Dict, Optional, Tuple, Type, TypeVar, Union
 
 import numpy as np
@@ -10,7 +8,7 @@
 from stable_baselines3.common.on_policy_algorithm import OnPolicyAlgorithm
 from stable_baselines3.common.policies import BasePolicy
 from stable_baselines3.common.type_aliases import GymEnv, MaybeCallback, Schedule
-from stable_baselines3.common.utils import explained_variance, get_schedule_fn, obs_as_tensor, safe_mean
+from stable_baselines3.common.utils import explained_variance, get_schedule_fn, obs_as_tensor
 from stable_baselines3.common.vec_env import VecEnv
 from torch.nn import functional as F
 
@@ -241,7 +239,7 @@ def collect_rollouts(
             if not callback.on_step():
                 return False
 
-            self._update_info_buffer(infos)
+            self._update_info_buffer(infos, dones)
             n_steps += 1
 
             if isinstance(self.action_space, spaces.Discrete):
@@ -463,17 +461,7 @@ def learn(  # type: ignore[override]
 
             # Display training infos
             if log_interval is not None and iteration % log_interval == 0:
-                assert self.ep_info_buffer is not None
-                time_elapsed = max((time.time_ns() - self.start_time) / 1e9, sys.float_info.epsilon)
-                fps = int((self.num_timesteps - self._num_timesteps_at_start) / time_elapsed)
-                self.logger.record("time/iterations", iteration, exclude="tensorboard")
-                if len(self.ep_info_buffer) > 0 and len(self.ep_info_buffer[0]) > 0:
-                    self.logger.record("rollout/ep_rew_mean", safe_mean([ep_info["r"] for ep_info in self.ep_info_buffer]))
-                    self.logger.record("rollout/ep_len_mean", safe_mean([ep_info["l"] for ep_info in self.ep_info_buffer]))
-                self.logger.record("time/fps", fps)
-                self.logger.record("time/time_elapsed", int(time_elapsed), exclude="tensorboard")
-                self.logger.record("time/total_timesteps", self.num_timesteps, exclude="tensorboard")
-                self.logger.dump(step=self.num_timesteps)
+                self._dump_logs(iteration)
 
             self.train()
 

sb3_contrib/ppo_recurrent/ppo_recurrent.py

@@ -1,5 +1,3 @@
-import sys
-import time
 from copy import deepcopy
 from typing import Any, ClassVar, Dict, Optional, Type, TypeVar, Union
 
@@ -11,7 +9,7 @@
 from stable_baselines3.common.on_policy_algorithm import OnPolicyAlgorithm
 from stable_baselines3.common.policies import BasePolicy
 from stable_baselines3.common.type_aliases import GymEnv, MaybeCallback, Schedule
-from stable_baselines3.common.utils import explained_variance, get_schedule_fn, obs_as_tensor, safe_mean
+from stable_baselines3.common.utils import explained_variance, get_schedule_fn, obs_as_tensor
 from stable_baselines3.common.vec_env import VecEnv
 
 from sb3_contrib.common.recurrent.buffers import RecurrentDictRolloutBuffer, RecurrentRolloutBuffer
@@ -260,7 +258,7 @@ def collect_rollouts(
             if not callback.on_step():
                 return False
 
-            self._update_info_buffer(infos)
+            self._update_info_buffer(infos, dones)
             n_steps += 1
 
             if isinstance(self.action_space, spaces.Discrete):
@@ -453,42 +451,11 @@ def learn(
         reset_num_timesteps: bool = True,
         progress_bar: bool = False,
     ) -> SelfRecurrentPPO:
-        iteration = 0
-
-        total_timesteps, callback = self._setup_learn(
-            total_timesteps,
-            callback,
-            reset_num_timesteps,
-            tb_log_name,
-            progress_bar,
+        return super().learn(
+            total_timesteps=total_timesteps,
+            callback=callback,
+            log_interval=log_interval,
+            tb_log_name=tb_log_name,
+            reset_num_timesteps=reset_num_timesteps,
+            progress_bar=progress_bar,
         )
-
-        callback.on_training_start(locals(), globals())
-
-        while self.num_timesteps < total_timesteps:
-            continue_training = self.collect_rollouts(self.env, callback, self.rollout_buffer, n_rollout_steps=self.n_steps)
-
-            if not continue_training:
-                break
-
-            iteration += 1
-            self._update_current_progress_remaining(self.num_timesteps, total_timesteps)
-
-            # Display training infos
-            if log_interval is not None and iteration % log_interval == 0:
-                time_elapsed = max((time.time_ns() - self.start_time) / 1e9, sys.float_info.epsilon)
-                fps = int((self.num_timesteps - self._num_timesteps_at_start) / time_elapsed)
-                self.logger.record("time/iterations", iteration, exclude="tensorboard")
-                if len(self.ep_info_buffer) > 0 and len(self.ep_info_buffer[0]) > 0:
-                    self.logger.record("rollout/ep_rew_mean", safe_mean([ep_info["r"] for ep_info in self.ep_info_buffer]))
-                    self.logger.record("rollout/ep_len_mean", safe_mean([ep_info["l"] for ep_info in self.ep_info_buffer]))
-                self.logger.record("time/fps", fps)
-                self.logger.record("time/time_elapsed", int(time_elapsed), exclude="tensorboard")
-                self.logger.record("time/total_timesteps", self.num_timesteps, exclude="tensorboard")
-                self.logger.dump(step=self.num_timesteps)
-
-            self.train()
-
-        callback.on_training_end()
-
-        return self

sb3_contrib/version.txt

@@ -1 +1 @@
-2.3.0a4
+2.3.0a5

setup.py

@@ -65,7 +65,7 @@
     packages=[package for package in find_packages() if package.startswith("sb3_contrib")],
     package_data={"sb3_contrib": ["py.typed", "version.txt"]},
     install_requires=[
-        "stable_baselines3>=2.3.0a4,<3.0",
+        "stable_baselines3>=2.3.0a5,<3.0",
     ],
     description="Contrib package of Stable Baselines3, experimental code.",
     author="Antonin Raffin",