make sure we can extend the wrappers config

scripts/configs/wrappers/base_wrappers.yaml

@@ -1,8 +1,7 @@
-- _target_: gymnasium.wrappers.TimeLimit
-  max_episode_steps: 1000
-
-- _target_: furuta.rl.wrappers.HistoryWrapper
-  steps: 2
-  use_continuity_cost: True
-
-- _target_: stable_baselines3.common.monitor.Monitor
+wrappers:
+  - _target_: gymnasium.wrappers.TimeLimit
+    max_episode_steps: 1000
+  - _target_: furuta.rl.wrappers.HistoryWrapper
+    steps: 2
+    use_continuity_cost: True
+  - _target_: stable_baselines3.common.monitor.Monitor

scripts/configs/wrappers/real_wrappers.yaml

@@ -1,8 +1,9 @@
 # import common wrappers between sim/real
 defaults:
-  - base_wrappers
+  - base_wrappers@_here_
 
-- _target_: furuta.rl.wrappers.GentlyTerminating
-- _target_: furuta.rl.wrappers.ControlFrequency
+wrappers:
+  - _target_: furuta.rl.wrappers.GentlyTerminating
+  - _target_: furuta.rl.wrappers.ControlFrequency
 # - _target_: furuta.rl.wrappers.MCAPLogger
 #   use_sim_time: False

scripts/train.py

@@ -47,7 +47,7 @@ def main(cfg: DictConfig):
     seed_everything(env, cfg.seed, cfg.cudnn_deterministic)
 
     # setup wrappers
-    for wrapper in cfg.wrappers:
+    for wrapper in cfg.wrappers.wrappers:
         env = hydra.utils.instantiate(wrapper, env=env)
 
     # don't paralelize if it's the real robot