Added script for evaluating saved models

evaluate.py

@@ -0,0 +1,56 @@
+import argparse
+import torch
+import time
+import numpy as np
+from pathlib import Path
+from torch.autograd import Variable
+from utils.make_env import make_env
+from algorithms.maddpg import MADDPG
+
+
+def run(config):
+    model_path = (Path('./models') / config.env_id / config.model_name /
+                  ('run%i' % config.run_num) / 'model.pt')
+
+    env = make_env(config.env_id)
+    maddpg = MADDPG.init_from_save(model_path)
+    ifi = 1 / config.fps  # inter-frame interval
+
+    for ep_i in range(config.n_episodes):
+        print("Episode %i of %i" % (ep_i + 1, config.n_episodes))
+        obs = env.reset()
+        env.render('human')
+        for t_i in range(config.episode_length):
+            calc_start = time.time()
+            # rearrange observations to be per agent, and convert to torch Variable
+            torch_obs = [Variable(torch.Tensor(obs[i]),
+                                  requires_grad=False)
+                         for i in range(maddpg.nagents)]
+            # get actions as torch Variables
+            torch_actions = maddpg.step(torch_obs, training=False)
+            # convert actions to numpy arrays
+            actions = [ac.data.numpy() for ac in torch_actions]
+            obs, rewards, dones, infos = env.step(actions)
+            calc_end = time.time()
+            elapsed = calc_end - calc_start
+            if elapsed < ifi:
+                time.sleep(ifi - elapsed)
+            env.render('human')
+
+    env.close()
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument("env_id", help="Name of environment")
+    parser.add_argument("model_name",
+                        help="Name of directory to store " +
+                             "model/training contents")
+    parser.add_argument("run_num", default=1, type=int)
+    parser.add_argument("--n_episodes", default=10, type=int)
+    parser.add_argument("--episode_length", default=100, type=int)
+    parser.add_argument("--fps", default=30, type=int)
+
+    config = parser.parse_args()
+
+    run(config)