Set render=True by default for panda gym envs (#311)

* correct visualization by adding render:True to panda_gym envs in absense of --no-render

* remove buried README note now that args are passed under the hood

* update changelog with changes for panda-gym visualization

* handle uninitialized env_kwargs

* add comment

* fix typo changelog

Co-authored-by: Quentin GALLOUÉDEC <gallouedec.quentin@gmail.com>
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>

Author: eric-y-chen

CHANGELOG.md

@@ -5,6 +5,7 @@
 ### New Features
 - Specifying custom policies in yaml file is now supported (@Rick-v-E)
 - Added ``monitor_kwargs`` parameter
+- Handle the `env_kwargs` of `render:True` under the hood for panda-gym v1 envs in `enjoy` replay to match visualzation behavior of other envs
 
 ### Bug fixes
 - Allow `python -m rl_zoo3.cli` to be called directly
@@ -239,7 +240,7 @@
 - Clarify n_evaluations help text (@jkterry)
 - Simplified hyperparameters files making use of defaults
 - Added new TQC+HER agents
-- Add `panda-gym`environments (@qgallouedec)
+- Add `panda-gym` environments (@qgallouedec)
 
 ## Release 1.0 (2021-03-17)
 

README.md

@@ -516,8 +516,6 @@ We used the v1 environments.
 |----------|-------------|-------------------|-----------|------------|------------|
 | HER+TQC | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |
 
-To visualize the result, you can pass `--env-kwargs render:True` to the enjoy script.
-
 
 ### MiniGrid Envs
 

rl_zoo3/exp_manager.py

@@ -509,6 +509,11 @@ def is_robotics_env(env_id: str) -> bool:
         entry_point = gym.envs.registry.env_specs[env_id].entry_point  # pytype: disable=module-attr
         return "gym.envs.robotics" in str(entry_point) or "panda_gym.envs" in str(entry_point)
 
+    @staticmethod
+    def is_panda_gym(env_id: str) -> bool:
+        entry_point = gym.envs.registry.env_specs[env_id].entry_point  # pytype: disable=module-attr
+        return "panda_gym.envs" in str(entry_point)
+
     def _maybe_normalize(self, env: VecEnv, eval_env: bool) -> VecEnv:
         """
         Wrap the env into a VecNormalize wrapper if needed

rl_zoo3/utils.py

@@ -231,6 +231,13 @@ def create_test_env(
         vec_env_cls = SubprocVecEnv
         # start_method = 'spawn' for thread safe
 
+    # panda-gym is based on pybullet, whose rendering requires to be configure at initialization
+    if ExperimentManager.is_panda_gym(env_id) and should_render:
+        if env_kwargs is None:
+            env_kwargs = {"render": True}
+        else:
+            env_kwargs["render"] = True
+
     env = make_vec_env(
         env_id,
         n_envs=n_envs,