From 7ce7b6a8c25d5862ec4850480191ac66ea02d3c8 Mon Sep 17 00:00:00 2001
From: Antonin RAFFIN <antonin.raffin@ensta.org>
Date: Sat, 18 Jun 2022 10:52:52 +0200
Subject: [PATCH] Update defaults for offpolicy algos with features extractor
 (#935)

---
 docs/misc/changelog.rst           |  4 +++-
 stable_baselines3/sac/policies.py | 11 ++++-------
 stable_baselines3/td3/policies.py |  8 ++++----
 stable_baselines3/version.txt     |  2 +-
 4 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/docs/misc/changelog.rst b/docs/misc/changelog.rst
index 1c88691..480b08e 100644
--- a/docs/misc/changelog.rst
+++ b/docs/misc/changelog.rst
@@ -4,7 +4,7 @@ Changelog
 ==========
 
 
-Release 1.5.1a8 (WIP)
+Release 1.5.1a9 (WIP)
 ---------------------------
 
 Breaking Changes:
@@ -12,6 +12,8 @@ Breaking Changes:
 - Changed the way policy "aliases" are handled ("MlpPolicy", "CnnPolicy", ...), removing the former
   ``register_policy`` helper, ``policy_base`` parameter and using ``policy_aliases`` static attributes instead (@Gregwar)
 - SB3 now requires PyTorch >= 1.11
+- Changed the default network architecture when using ``CnnPolicy`` or ``MultiInputPolicy`` with SAC or DDPG/TD3,
+  ``share_features_extractor`` is now set to False by default and the ``net_arch=[256, 256]`` (instead of ``net_arch=[]`` that was before)
 
 New Features:
 ^^^^^^^^^^^^^
diff --git a/stable_baselines3/sac/policies.py b/stable_baselines3/sac/policies.py
index 6fcbea1..255bd75 100644
--- a/stable_baselines3/sac/policies.py
+++ b/stable_baselines3/sac/policies.py
@@ -235,7 +235,7 @@ class SACPolicy(BasePolicy):
         optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
         optimizer_kwargs: Optional[Dict[str, Any]] = None,
         n_critics: int = 2,
-        share_features_extractor: bool = True,
+        share_features_extractor: bool = False,
     ):
         super().__init__(
             observation_space,
@@ -248,10 +248,7 @@ class SACPolicy(BasePolicy):
         )
 
         if net_arch is None:
-            if features_extractor_class == NatureCNN:
-                net_arch = []
-            else:
-                net_arch = [256, 256]
+            net_arch = [256, 256]
 
         actor_arch, critic_arch = get_actor_critic_arch(net_arch)
 
@@ -422,7 +419,7 @@ class CnnPolicy(SACPolicy):
         optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
         optimizer_kwargs: Optional[Dict[str, Any]] = None,
         n_critics: int = 2,
-        share_features_extractor: bool = True,
+        share_features_extractor: bool = False,
     ):
         super().__init__(
             observation_space,
@@ -493,7 +490,7 @@ class MultiInputPolicy(SACPolicy):
         optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
         optimizer_kwargs: Optional[Dict[str, Any]] = None,
         n_critics: int = 2,
-        share_features_extractor: bool = True,
+        share_features_extractor: bool = False,
     ):
         super().__init__(
             observation_space,
diff --git a/stable_baselines3/td3/policies.py b/stable_baselines3/td3/policies.py
index f3ed530..8781b32 100644
--- a/stable_baselines3/td3/policies.py
+++ b/stable_baselines3/td3/policies.py
@@ -119,7 +119,7 @@ class TD3Policy(BasePolicy):
         optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
         optimizer_kwargs: Optional[Dict[str, Any]] = None,
         n_critics: int = 2,
-        share_features_extractor: bool = True,
+        share_features_extractor: bool = False,
     ):
         super().__init__(
             observation_space,
@@ -134,7 +134,7 @@ class TD3Policy(BasePolicy):
         # Default network architecture, from the original paper
         if net_arch is None:
             if features_extractor_class == NatureCNN:
-                net_arch = []
+                net_arch = [256, 256]
             else:
                 net_arch = [400, 300]
 
@@ -281,7 +281,7 @@ class CnnPolicy(TD3Policy):
         optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
         optimizer_kwargs: Optional[Dict[str, Any]] = None,
         n_critics: int = 2,
-        share_features_extractor: bool = True,
+        share_features_extractor: bool = False,
     ):
         super().__init__(
             observation_space,
@@ -335,7 +335,7 @@ class MultiInputPolicy(TD3Policy):
         optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
         optimizer_kwargs: Optional[Dict[str, Any]] = None,
         n_critics: int = 2,
-        share_features_extractor: bool = True,
+        share_features_extractor: bool = False,
     ):
         super().__init__(
             observation_space,
diff --git a/stable_baselines3/version.txt b/stable_baselines3/version.txt
index 511e75b..125ec27 100644
--- a/stable_baselines3/version.txt
+++ b/stable_baselines3/version.txt
@@ -1 +1 @@
-1.5.1a8
+1.5.1a9