[Feature] Simplify extending benchmarl and improve extending examples (…

…#89) * amend * empty * amend * amend * amend * amend * amend * amend * change examples * amend * copyright * amend * amend * amend * amend * amend * amend * amend * amend * empty * empty
facebookresearch · Jun 10, 2024 · 87b4f3e · 87b4f3e
1 parent 61c04f7
commit 87b4f3e
Show file tree

Hide file tree

Showing 57 changed files with 250 additions and 204 deletions.
diff --git a/README.md b/README.md
@@ -253,7 +253,7 @@ They differ based on many aspects, here is a table with the current environments
 > BenchMARL uses the [TorchRL MARL API](https://github.com/pytorch/rl/issues/1463) for grouping agents.
 > In competitive environments like MPE, for example, teams will be in different groups. Each group has its own loss,
 > models, buffers, and so on. Parameter sharing options refer to sharing within the group. See the example on [creating
-> a custom algorithm](examples/extending/algorithm/custom_algorithm.py) for more info.
+> a custom algorithm](examples/extending/algorithm/algorithms/customalgorithm.py) for more info.
 
 **Models**. Models are neural networks used to process data. They can be used as actors (policies) or, 
 when requested, as critics. We provide a set of base models (layers) and a SequenceModel to concatenate

diff --git a/benchmarl/__init__.py b/benchmarl/__init__.py
@@ -35,6 +35,6 @@ def _load_hydra_schemas():
             cs.store(name=f"{algo_name}_config", group="algorithm", node=algo_schema)
         # Load task schemas
         for task_schema_name, task_schema in _task_class_registry.items():
-            cs.store(name=task_schema_name, group="task", node=task_schema)
+            cs.store(name=f"{task_schema_name}_config", group="task", node=task_schema)
 
     _load_hydra_schemas()
diff --git a/benchmarl/algorithms/common.py b/benchmarl/algorithms/common.py
@@ -358,14 +358,15 @@ def get_from_yaml(cls, path: Optional[str] = None):
 
         Returns: the loaded AlgorithmConfig
         """
+
         if path is None:
-            return cls(
-                **AlgorithmConfig._load_from_yaml(
-                    name=cls.associated_class().__name__,
-                )
+            config = AlgorithmConfig._load_from_yaml(
+                name=cls.associated_class().__name__
             )
+
         else:
-            return cls(**_read_yaml_config(path))
+            config = _read_yaml_config(path)
+        return cls(**config)
 
     @staticmethod
     @abstractmethod

diff --git a/benchmarl/conf/task/pettingzoo/multiwalker.yaml b/benchmarl/conf/task/pettingzoo/multiwalker.yaml
@@ -1,6 +1,6 @@
 defaults:
-  - _self_
   - pettingzoo_multiwalker_config
+  - _self_
 
 task: "multiwalker_v9"
 # number of bipedal walker agents in environment

diff --git a/benchmarl/conf/task/pettingzoo/simple_adversary.yaml b/benchmarl/conf/task/pettingzoo/simple_adversary.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_adversary_config
+  - _self_
+
 
 task: "simple_adversary_v3"
 N: 2

diff --git a/benchmarl/conf/task/pettingzoo/simple_crypto.yaml b/benchmarl/conf/task/pettingzoo/simple_crypto.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_crypto_config
+  - _self_
+
 
 task: "simple_crypto_v3"
 max_cycles: 100
diff --git a/benchmarl/conf/task/pettingzoo/simple_push.yaml b/benchmarl/conf/task/pettingzoo/simple_push.yaml
@@ -1,6 +1,7 @@
 defaults:
+  -  pettingzoo_simple_push_config
   - _self_
-  - pettingzoo_simple_push_config
+
 
 task: "simple_push_v3"
 max_cycles: 100
diff --git a/benchmarl/conf/task/pettingzoo/simple_reference.yaml b/benchmarl/conf/task/pettingzoo/simple_reference.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_reference_config
+  - _self_
+
 
 task: "simple_reference_v3"
 max_cycles: 100

diff --git a/benchmarl/conf/task/pettingzoo/simple_speaker_listener.yaml b/benchmarl/conf/task/pettingzoo/simple_speaker_listener.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_speaker_listener_config
+  - _self_
+
 
 task: "simple_speaker_listener_v4"
 max_cycles: 100
diff --git a/benchmarl/conf/task/pettingzoo/simple_spread.yaml b/benchmarl/conf/task/pettingzoo/simple_spread.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_spread_config
+  - _self_
+
 
 task: "simple_spread_v3"
 max_cycles: 100

diff --git a/benchmarl/conf/task/pettingzoo/simple_tag.yaml b/benchmarl/conf/task/pettingzoo/simple_tag.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_tag_config
+  - _self_
+
 
 task: "simple_tag_v3"
 num_good: 2

diff --git a/benchmarl/conf/task/pettingzoo/simple_world_comm.yaml b/benchmarl/conf/task/pettingzoo/simple_world_comm.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_simple_world_comm_config
+  - _self_
+
 
 task: "simple_world_comm_v3"
 num_good: 2

diff --git a/benchmarl/conf/task/pettingzoo/waterworld.yaml b/benchmarl/conf/task/pettingzoo/waterworld.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - pettingzoo_waterworld_config
+  - _self_
+
 
 task: "waterworld_v4"
 max_cycles: 500

diff --git a/benchmarl/conf/task/vmas/balance.yaml b/benchmarl/conf/task/vmas/balance.yaml
@@ -1,7 +1,6 @@
 defaults:
-  - _self_
   - vmas_balance_config
-
+  - _self_
 
 max_steps: 100
 n_agents: 4

diff --git a/benchmarl/conf/task/vmas/ball_passage.yaml b/benchmarl/conf/task/vmas/ball_passage.yaml
@@ -1,6 +1,6 @@
 defaults:
-  - _self_
   - vmas_ball_passage_config
+  - _self_
 
 max_steps: 500
 n_passages: 1

diff --git a/benchmarl/conf/task/vmas/ball_trajectory.yaml b/benchmarl/conf/task/vmas/ball_trajectory.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_ball_trajectory_config
+  - _self_
+
 
 max_steps: 100
 joints: True
diff --git a/benchmarl/conf/task/vmas/buzz_wire.yaml b/benchmarl/conf/task/vmas/buzz_wire.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_buzz_wire_config
+  - _self_
+
 
 max_steps: 100
 random_start_angle: True

diff --git a/benchmarl/conf/task/vmas/discovery.yaml b/benchmarl/conf/task/vmas/discovery.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_discovery_config
+  - _self_
+
 
 max_steps: 100
 n_agents: 5

diff --git a/benchmarl/conf/task/vmas/dispersion.yaml b/benchmarl/conf/task/vmas/dispersion.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_dispersion_config
+  - _self_
+
 
 
 max_steps: 100

diff --git a/benchmarl/conf/task/vmas/dropout.yaml b/benchmarl/conf/task/vmas/dropout.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_dropout_config
+  - _self_
+
 
 
 max_steps: 100

diff --git a/benchmarl/conf/task/vmas/flocking.yaml b/benchmarl/conf/task/vmas/flocking.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_flocking_config
+  - _self_
+
 
 max_steps: 100
 n_agents: 4

diff --git a/benchmarl/conf/task/vmas/give_way.yaml b/benchmarl/conf/task/vmas/give_way.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_give_way_config
+  - _self_
+
 
 max_steps: 100
 mirror_passage: False

diff --git a/benchmarl/conf/task/vmas/joint_passage.yaml b/benchmarl/conf/task/vmas/joint_passage.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_joint_passage_config
+  - _self_
+
 
 max_steps: 500
 n_passages: 1

diff --git a/benchmarl/conf/task/vmas/joint_passage_size.yaml b/benchmarl/conf/task/vmas/joint_passage_size.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_joint_passage_size_config
+  - _self_
+
 
 max_steps: 500
 n_passages: 3

diff --git a/benchmarl/conf/task/vmas/multi_give_way.yaml b/benchmarl/conf/task/vmas/multi_give_way.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_multi_give_way_config
+  - _self_
+
 
 max_steps: 200
 agent_collision_penalty: -0.1
diff --git a/benchmarl/conf/task/vmas/navigation.yaml b/benchmarl/conf/task/vmas/navigation.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_navigation_config
+  - _self_
+
 
 
 max_steps: 100

diff --git a/benchmarl/conf/task/vmas/passage.yaml b/benchmarl/conf/task/vmas/passage.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_passage_config
+  - _self_
+
 
 max_steps: 500
 n_passages: 1

diff --git a/benchmarl/conf/task/vmas/reverse_transport.yaml b/benchmarl/conf/task/vmas/reverse_transport.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_reverse_transport_config
+  - _self_
+
 
 max_steps: 100
 n_agents: 4

diff --git a/benchmarl/conf/task/vmas/sampling.yaml b/benchmarl/conf/task/vmas/sampling.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_sampling_config
+  - _self_
+
 
 
 max_steps: 100

diff --git a/benchmarl/conf/task/vmas/simple_adversary.yaml b/benchmarl/conf/task/vmas/simple_adversary.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_simple_adversary_config
+  - _self_
+
 
 max_steps: 100
 n_agents: 3

diff --git a/benchmarl/conf/task/vmas/simple_crypto.yaml b/benchmarl/conf/task/vmas/simple_crypto.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_simple_crypto_config
+  - _self_
+
 
 max_steps: 100
 dim_c: 4
diff --git a/benchmarl/conf/task/vmas/simple_push.yaml b/benchmarl/conf/task/vmas/simple_push.yaml
@@ -1,5 +1,6 @@
 defaults:
-  - _self_
   - vmas_simple_push_config
+  - _self_
+
 
 max_steps: 100
diff --git a/benchmarl/conf/task/vmas/simple_reference.yaml b/benchmarl/conf/task/vmas/simple_reference.yaml
@@ -1,5 +1,6 @@
 defaults:
-  - _self_
   - vmas_simple_reference_config
+  - _self_
+
 
 max_steps: 100
diff --git a/benchmarl/conf/task/vmas/simple_speaker_listener.yaml b/benchmarl/conf/task/vmas/simple_speaker_listener.yaml
@@ -1,5 +1,6 @@
 defaults:
-  - _self_
   - vmas_simple_speaker_listener_config
+  - _self_
+
 
 max_steps: 100
diff --git a/benchmarl/conf/task/vmas/simple_spread.yaml b/benchmarl/conf/task/vmas/simple_spread.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_simple_spread_config
+  - _self_
+
 
 max_steps: 100
 n_agents: 3

diff --git a/benchmarl/conf/task/vmas/simple_tag.yaml b/benchmarl/conf/task/vmas/simple_tag.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_simple_tag_config
+  - _self_
+
 
 max_steps: 100
 num_good_agents: 1

diff --git a/benchmarl/conf/task/vmas/simple_world_comm.yaml b/benchmarl/conf/task/vmas/simple_world_comm.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_simple_world_comm_config
+  - _self_
+
 
 max_steps: 100
 num_good_agents: 2

diff --git a/benchmarl/conf/task/vmas/transport.yaml b/benchmarl/conf/task/vmas/transport.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_transport_config
+  - _self_
+
 
 
 max_steps: 100

diff --git a/benchmarl/conf/task/vmas/wheel.yaml b/benchmarl/conf/task/vmas/wheel.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_wheel_config
+  - _self_
+
 
 max_steps: 100
 n_agents: 4

diff --git a/benchmarl/conf/task/vmas/wind_flocking.yaml b/benchmarl/conf/task/vmas/wind_flocking.yaml
@@ -1,6 +1,7 @@
 defaults:
-  - _self_
   - vmas_wind_flocking_config
+  - _self_
+
 
 max_steps: 100
 horizon: 100