From 7de41d2db0e5eca31107312d405e20ff3a7da39e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Edgar=20Du=C3=A9=C3=B1ez-Guzm=C3=A1n?= <duenez@google.com>
Date: Mon, 6 Mar 2023 05:18:10 -0800
Subject: [PATCH] Use named policies, rather than the default "av" in
 `view_models`.

Fixes https://github.com/deepmind/meltingpot/issues/110

PiperOrigin-RevId: 514378189
Change-Id: I33c69386262299355164ec5671414cc501762e9b
---
 examples/rllib/view_models.py | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/examples/rllib/view_models.py b/examples/rllib/view_models.py
index a131a526..08e7b62c 100644
--- a/examples/rllib/view_models.py
+++ b/examples/rllib/view_models.py
@@ -60,8 +60,10 @@ def main():
   # Create a new environment to visualise
   env = utils.env_creator(config["env_config"]).get_dmlab2d_env()
 
-  num_bots = config["env_config"]["num_players"]
-  bots = [utils.RayModelPolicy(trainer, "av")] * num_bots
+  bots = [
+      utils.RayModelPolicy(trainer, f"agent_{i}")
+      for i in range(len(config["env_config"]["default_player_roles"]))
+  ]
 
   timestep = env.reset()
   states = [bot.initial_state() for bot in bots]