Multiprocessing support for off policy algorithms (#50)

* TQC support for multienv

* Add optional layer norm for TQC

* Add layer nprm for all policies

* Revert "Add layer nprm for all policies"

This reverts commit 1306c3c64eb12613464982c66cb416a3bbc66285.

* Revert "Add optional layer norm for TQC"

This reverts commit 200222e3a8878007aa6032d540ae74274a4d0788.

* Add experimental support to train off-policy algorithms with multiple envs

* Bump version

* Update version

docs/guide/algos.rst

@@ -8,8 +8,8 @@ along with some useful characteristics: support for discrete/continuous actions,
 ============ =========== ============ ================= =============== ================
 Name         ``Box``     ``Discrete`` ``MultiDiscrete`` ``MultiBinary`` Multi Processing
 ============ =========== ============ ================= =============== ================
-TQC          ✔️          ❌            ❌                ❌              ❌
+TQC          ✔️          ❌            ❌                ❌                ✔️
-QR-DQN       ️❌          ️✔️            ❌                ❌              ❌
+QR-DQN       ️❌          ️✔️            ❌                ❌                ✔️ 
 ============ =========== ============ ================= =============== ================
 
 

docs/misc/changelog.rst

@@ -4,15 +4,16 @@ Changelog
 ==========
 
 
-Release 1.3.1a3 (WIP)
+Release 1.3.1a4 (WIP)
 -------------------------------
 
 Breaking Changes:
 ^^^^^^^^^^^^^^^^^
-- Upgraded to Stable-Baselines3 >= 1.3.1a3
+- Upgraded to Stable-Baselines3 >= 1.3.1a4
 
 New Features:
 ^^^^^^^^^^^^^
+- Added experimental support to train off-policy algorithms with multiple envs (note: ``HerReplayBuffer`` currently not supported)
 
 Bug Fixes:
 ^^^^^^^^^^

docs/modules/qrdqn.rst

@@ -33,7 +33,7 @@ Can I use?
 ----------
 
 -  Recurrent policies: ❌
--  Multi processing: ❌
+-  Multi processing: ✔️
 -  Gym spaces:
 
 

docs/modules/tqc.rst

@@ -32,7 +32,7 @@ Can I use?
 ----------
 
 -  Recurrent policies: ❌
--  Multi processing: ❌
+-  Multi processing: ✔️
 -  Gym spaces:
 
 

sb3_contrib/qrdqn/qrdqn.py

@@ -110,6 +110,7 @@ class QRDQN(OffPolicyAlgorithm):
             sde_support=False,
             optimize_memory_usage=optimize_memory_usage,
             supported_action_spaces=(gym.spaces.Discrete,),
+            support_multi_env=True,
         )
 
         self.exploration_initial_eps = exploration_initial_eps

sb3_contrib/tqc/tqc.py

@@ -122,6 +122,7 @@ class TQC(OffPolicyAlgorithm):
             use_sde_at_warmup=use_sde_at_warmup,
             optimize_memory_usage=optimize_memory_usage,
             supported_action_spaces=(gym.spaces.Box),
+            support_multi_env=True,
         )
 
         self.target_entropy = target_entropy

sb3_contrib/version.txt

@@ -1 +1 @@
-1.3.1a3
+1.3.1a4

setup.py

@@ -63,7 +63,7 @@ setup(
     packages=[package for package in find_packages() if package.startswith("sb3_contrib")],
     package_data={"sb3_contrib": ["py.typed", "version.txt"]},
     install_requires=[
-        "stable_baselines3>=1.3.1a3",
+        "stable_baselines3>=1.3.1a4",
     ],
     description="Contrib package of Stable Baselines3, experimental code.",
     author="Antonin Raffin",

tests/test_run.py

@@ -1,4 +1,6 @@
+import gym
 import pytest
+from stable_baselines3.common.env_util import make_vec_env
 
 from sb3_contrib import QRDQN, TQC
 
@@ -56,3 +58,31 @@ def test_qrdqn():
         create_eval_env=True,
     )
     model.learn(total_timesteps=500, eval_freq=250)
+
+
+@pytest.mark.parametrize("model_class", [TQC, QRDQN])
+def test_offpolicy_multi_env(model_class):
+    if model_class in [TQC]:
+        env_id = "Pendulum-v0"
+        policy_kwargs = dict(net_arch=[64], n_critics=1)
+    else:
+        env_id = "CartPole-v1"
+        policy_kwargs = dict(net_arch=[64])
+
+    def make_env():
+        env = gym.make(env_id)
+        # to check that the code handling timeouts runs
+        env = gym.wrappers.TimeLimit(env, 50)
+        return env
+
+    env = make_vec_env(make_env, n_envs=2)
+    model = model_class(
+        "MlpPolicy",
+        env,
+        policy_kwargs=policy_kwargs,
+        learning_starts=100,
+        buffer_size=10000,
+        verbose=0,
+        train_freq=5,
+    )
+    model.learn(total_timesteps=150)