test experts

tests/test_expert.py

@@ -0,0 +1,60 @@
+import gpudrive
+import torch
+import pytest
+
+@pytest.fixture(scope="module")
+def simulation_results():
+    # Create an instance of RewardParams
+    reward_params = gpudrive.RewardParams()
+    reward_params.rewardType = gpudrive.RewardType.DistanceBased  # Or any other value from the enum
+    reward_params.distanceToGoalThreshold = 1.0  # Set appropriate values
+    reward_params.distanceToExpertThreshold = 1.0  # Set appropriate values
+
+    # Create an instance of Parameters
+    params = gpudrive.Parameters()
+    params.polylineReductionThreshold = 0.5  # Set appropriate value
+    params.observationRadius = 10.0  # Set appropriate value
+    params.collisionBehaviour = gpudrive.CollisionBehaviour.AgentStop  # Set appropriate value
+    params.datasetInitOptions = gpudrive.DatasetInitOptions.PadN  # Set appropriate value
+    params.rewardParams = reward_params  # Set the rewardParams attribute to the instance created above
+    params.maxNumControlledVehicles = 0
+    params.IgnoreNonVehicles = True
+
+    # Now use the 'params' instance when creating SimManager
+    sim = gpudrive.SimManager(
+        exec_mode=gpudrive.madrona.ExecMode.CPU,
+        gpu_id=0,
+        num_worlds=2,
+        auto_reset=False,
+        json_path="nocturne_data",
+        params=params,
+        enable_batch_renderer=False, # Optional parameter
+        batch_render_view_width=1024,
+        batch_render_view_height=1024
+    )
+
+    done = sim.done_tensor().to_torch()
+    info = sim.info_tensor().to_torch()
+    shape = sim.shape_tensor().to_torch()
+
+    while not done.all():
+        sim.step()
+
+    return sim
+
+def test_goal_reaching(simulation_results):
+    sim = simulation_results
+    info, shape = sim.info_tensor().to_torch(), sim.shape_tensor().to_torch()
+    info_valid = info[info[:, :, -1] == float(gpudrive.EntityType.Vehicle)]
+    goal_reached = info_valid[:, -2].sum().item()
+    assert goal_reached == shape[:, 0].sum().item()
+    print("Test passed!")
+
+def test_collision_rate(simulation_results):
+    sim = simulation_results
+    info = sim.info_tensor().to_torch()
+
+    info_valid = info[info[:, :, -1] == float(gpudrive.EntityType.Vehicle)]
+    collisions = info_valid[:, :3].sum().item()
+    assert collisions == 0
+    print("Test passed!")