[Tutorial] PettingZoo Parallel competitive tutorial (#2047)

Co-authored-by: Vincent Moens <vmoens@meta.com>
Co-authored-by: Vincent Moens <vincentmoens@gmail.com>

docs/requirements.txt

@@ -25,4 +25,5 @@ imageio[ffmpeg,pyav]
 memory_profiler
 pyrender
 pytest
-vmas==1.2.11
+vmas
+pettingzoo[mpe]==1.24.3

docs/source/index.rst

@@ -111,6 +111,7 @@ Advanced
 .. toctree::
    :maxdepth: 1
 
+   tutorials/multiagent_competitive_ddpg
    tutorials/multi_task
    tutorials/coding_ddpg
    tutorials/coding_dqn

torchrl/envs/libs/vmas.py

@@ -302,7 +302,7 @@ def _get_default_group_map(self, agent_names: List[str]):
             group_map = MarlGroupMapType.ALL_IN_ONE_GROUP.get_group_map(agent_names)
 
         # For BC-compatibility rename the "agent" group to "agents"
-        if "agent" in group_map:
+        if "agent" in group_map and len(group_map) == 1:
             agent_group = group_map["agent"]
             group_map["agents"] = agent_group
             del group_map["agent"]

torchrl/record/loggers/csv.py

@@ -9,6 +9,7 @@
 from pathlib import Path
 from typing import Dict, Optional, Sequence, Union
 
+import tensordict.utils
 import torch
 
 from tensordict import MemoryMappedTensor
@@ -196,3 +197,7 @@ def __repr__(self) -> str:
 
     def log_histogram(self, name: str, data: Sequence, **kwargs):
         raise NotImplementedError("Logging histograms in cvs is not permitted.")
+
+    def print_log_dir(self):
+        """Prints the log directory content."""
+        tensordict.utils.print_directory_tree(self.log_dir)