fix requirements

README.md

@@ -99,6 +99,8 @@ export PYTHONPATH="${CARLA_ROOT}/PythonAPI/carla":${PYTHONPATH}
 Install the package using flit. The ``--symlink`` flag is used to create a symlink to the package in the Python environment, so that changes to the package are immediately available without reinstallation. (``--pth-file`` also works, as an alternative to ``--symlink``.)
 
 ```bash
+conda create python=3.10 --name cardreamer
+conda activate cardreamer
 pip install flit
 flit install --symlink
 ```

docs/source/usage.rst

@@ -23,6 +23,8 @@ Then, install the package using flit. The ``--symlink`` flag is used to create a
 
 .. code-block:: console
 
+   conda create python=3.10 --name cardreamer
+   conda activate cardreamer
    pip install flit
    flit install --symlink
 
@@ -58,7 +60,7 @@ After creating the task, the visualization is automatically started if ``env.dis
    import car_dreamer
    import time
 
-   task, _ = car_dreamer.create_task('carla_four_lane', argv)
+   task, _ = car_dreamer.create_task('carla_four_lane')
    task.reset()
    while True:
       _, _, is_terminal, _ = task.step(12)  # 12 is the action of going straight and accelerating

dreamerv3/eval.py

@@ -108,9 +108,9 @@ def main(argv=None):
     config = embodied.Config({'dreamerv3': model_configs['defaults']})
     config = config.update({'dreamerv3': model_configs['small']})
 
-    parsed, other = embodied.Flags(scenario=['carla_navigation']).parse_known(argv)
-    for name in parsed.scenario:
-        print("Using scenario: ", name)
+    parsed, other = embodied.Flags(task=['carla_navigation']).parse_known(argv)
+    for name in parsed.task:
+        print("Using task: ", name)
         env, env_config = car_dreamer.create_task(name, argv)
         config = config.update(env_config)
     config = embodied.Flags(config).parse(other)

dreamerv3/eval_stats.py

@@ -4,54 +4,7 @@
 import argparse
 from pathlib import Path
 
-
-def compute_metrics_distance_based(episode_stats, task_success_distance_thresholds, task):
-    metrics = {}
-
-    travel_distances = np.array(episode_stats['sum_travel_distance'])
-    destination_reached = np.array(episode_stats['sum_destination_reached'])
-    time_exceeded = np.array(episode_stats['sum_time_exceeded'])
-    is_collision = np.array(episode_stats['sum_is_collision'])
-    out_of_lane = np.array(episode_stats['sum_out_of_lane'])
-    ttc = np.array(episode_stats['mean_ttc'])
-    wpt_dis = np.array(episode_stats['mean_wpt_dis'])
-    speed_norm = np.array(episode_stats['mean_speed_norm'])
-    lengths = np.array(episode_stats['lengths'])
-    scores = np.array(episode_stats['scores'])
-
-    num_episodes = len(travel_distances)
-    task_threshold = task_success_distance_thresholds.get(task, 0)
-
-    success_mask = travel_distances >= task_threshold
-    success_rate = np.mean(success_mask)
-
-    # Filter out failures after the tasks are successful
-    # is_collision[success_mask] = False
-    # time_exceeded[success_mask] = False
-    #out_of_lane[success_mask] = False
-
-    metrics['task'] = task
-    metrics['num_episodes'] = num_episodes
-    metrics['success_rate'] = success_rate
-    metrics['avg_travel_distance'] = np.mean(travel_distances)
-    metrics['avg_destination_reached'] = np.mean(destination_reached)
-    metrics['avg_destination_reached_ratio'] = np.sum(destination_reached) / num_episodes
-    metrics['avg_time_exceeded'] = np.mean(time_exceeded)
-    metrics['avg_time_exceeded_ratio'] = np.sum(time_exceeded) / num_episodes
-    metrics['avg_is_collision'] = np.mean(is_collision)
-    metrics['avg_is_collision_ratio'] = np.sum(is_collision) / num_episodes
-    metrics['avg_out_of_lane'] = np.mean(out_of_lane)
-    metrics['avg_out_of_lane_ratio'] = np.sum(out_of_lane) / num_episodes
-    metrics['avg_ttc'] = np.mean(ttc)
-    metrics['avg_wpt_dis'] = np.mean(wpt_dis)
-    metrics['avg_speed_norm'] = np.mean(speed_norm)
-    metrics['avg_length'] = np.mean(lengths)
-    metrics['avg_score'] = np.mean(scores)
-
-    return metrics
-
-
-def compute_metrics_destination_based(episode_stats, task):
+def compute_metrics_destination_based(episode_stats):
     metrics = {}
 
     travel_distances = np.array(episode_stats['sum_travel_distance'])
@@ -74,7 +27,6 @@ def compute_metrics_destination_based(episode_stats, task):
     time_exceeded[destination_reached > 0] = False
     out_of_lane[destination_reached > 0] = False
 
-    metrics['task'] = task
     metrics['num_episodes'] = num_episodes
     metrics['success_rate'] = success_rate
     metrics['avg_travel_distance'] = np.mean(travel_distances)
@@ -96,24 +48,8 @@ def compute_metrics_destination_based(episode_stats, task):
 
 
 def main(args):
-    task_success_distance_thresholds = {
-        'carla_left_turn_simple': 43,
-        'carla_left_turn_medium': 43,
-        'carla_left_turn_hard': 43,
-        'carla_right_turn_simple': 32,
-        'carla_right_turn_medium': 32,
-        'carla_right_turn_hard': 32,
-        'carla_navigation': 50,
-        'carla_lane_merge': 92,
-        'carla_four_lane': 70,
-        'carla_roundabout': 60,
-        'carla_overtake': 20,
-        'carla_message': 115,
-    }
 
     jsonl_file = Path(args.logdir) / 'metrics.jsonl'
-    task = args.task
-    method = args.method
 
     episode_stats = {
         'sum_travel_distance': [],
@@ -152,15 +88,9 @@ def main(args):
             if 'episode/score' in data:
                 episode_stats['scores'].append(data['episode/score'])
 
-    if method == 'distance':
-        final_metrics = compute_metrics_distance_based(episode_stats, task_success_distance_thresholds, task)
-    elif method == 'destination':
-        final_metrics = compute_metrics_destination_based(episode_stats, task)
-    else:
-        raise ValueError("Method should be either 'distance' or 'destination'")
+    final_metrics = compute_metrics_destination_based(episode_stats)
 
     table = [
-        ["Task", final_metrics['task']],
         ["Number of Episodes", final_metrics['num_episodes']],
         ["Success Rate", f"{final_metrics['success_rate']:.2%}"],
         ["Avg. Travel Distance", f"{final_metrics['avg_travel_distance']:.2f}"],

dreamerv3/requirements.txt

@@ -1,13 +1,13 @@
 cloudpickle==3.0.0
 crafter==1.8.3
-gym==0.26.2
+gym==0.12.5
 dm_control==1.0.18
 jax==0.4.26
 jaxlib==0.4.26+cuda12.cudnn89
 numpy==1.24.3
 optax==0.2.2
 rich==13.7.1
-ruamel.yaml==0.18.5
+ruamel.yaml==0.18.6
 tensorflow[and-cuda]==2.13.0
 tensorflow_probability==0.21.0
 scikit-image==0.23.2

pyproject.toml

@@ -20,4 +20,5 @@ dependencies = [
     "typing_extensions==4.5.0",
     "networkx==3.1",
     "Shapely==1.8.4",
+    "pyyaml==6.0.1",
 ]