Update planner.py

model/planner.py

@@ -13,13 +13,13 @@ def __init__(self, trajectory_len, feature_len, device, vectorize=True, test=Fal
         control_variables = th.Vector(dof=100, name="control_variables")
 
         # define prediction variable
-        predictions = th.Variable(data=torch.empty(1, 10, trajectory_len, 3), name="predictions")
+        predictions = th.Variable(torch.empty(1, 10, trajectory_len, 3), name="predictions")
 
         # define ref_line_info
-        ref_line_info = th.Variable(data=torch.empty(1, 1200, 5), name="ref_line_info")
+        ref_line_info = th.Variable(torch.empty(1, 1200, 5), name="ref_line_info")
 
         # define current state
-        current_state = th.Variable(data=torch.empty(1, 11, 8), name="current_state")
+        current_state = th.Variable(torch.empty(1, 11, 8), name="current_state")
 
         # set up objective
         objective = th.Objective()
@@ -66,34 +66,34 @@ def bicycle_model(control, current_state):
 
 # cost functions
 def acceleration(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
+    control = optim_vars[0].tensor.view(-1, 50, 2)
     acc = control[:, :, 0]
 
     return acc
 
 def jerk(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
+    control = optim_vars[0].tensor.view(-1, 50, 2)
     acc = control[:, :, 0]
     jerk = torch.diff(acc) / 0.1
 
     return jerk
 
 def steering(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
+    control = optim_vars[0].tensor.view(-1, 50, 2)
     steering = control[:, :, 1]
 
     return steering 
 
 def steering_change(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
+    control = optim_vars[0].tensor.view(-1, 50, 2)
     steering = control[:, :, 1]
     steering_change = torch.diff(steering) / 0.1
 
     return steering_change
 
 def speed(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
-    current_state = aux_vars[1].data[:, 0]
+    control = optim_vars[0].tensor.view(-1, 50, 2)
+    current_state = aux_vars[1].tensor[:, 0]
     velocity = torch.hypot(current_state[:, 3], current_state[:, 4]) 
     dt = 0.1
 
@@ -108,9 +108,9 @@ def speed(optim_vars, aux_vars):
 def lane_xy(optim_vars, aux_vars):
     global ref_points
 
-    control = optim_vars[0].data.view(-1, 50, 2)
-    ref_line = aux_vars[0].data
-    current_state = aux_vars[1].data[:, 0]
+    control = optim_vars[0].tensor.view(-1, 50, 2)
+    ref_line = aux_vars[0].tensor
+    current_state = aux_vars[1].tensor[:, 0]
 
     traj = bicycle_model(control, current_state)
     distance_to_ref = torch.cdist(traj[:, :, :2], ref_line[:, :, :2])
@@ -121,8 +121,8 @@ def lane_xy(optim_vars, aux_vars):
     return lane_error
 
 def lane_theta(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
-    current_state = aux_vars[1].data[:, 0]
+    control = optim_vars[0].tensor.view(-1, 50, 2)
+    current_state = aux_vars[1].tensor[:, 0]
 
     traj = bicycle_model(control, current_state)
     theta = traj[:, :, 2]
@@ -131,9 +131,9 @@ def lane_theta(optim_vars, aux_vars):
     return lane_error
 
 def red_light_violation(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
-    current_state = aux_vars[1].data[:, 0]
-    ref_line = aux_vars[0].data
+    control = optim_vars[0].tensor.view(-1, 50, 2)
+    current_state = aux_vars[1].tensor[:, 0]
+    ref_line = aux_vars[0].tensor
     red_light = ref_line[..., -1]
     dt = 0.1
 
@@ -150,10 +150,10 @@ def red_light_violation(optim_vars, aux_vars):
     return red_light_error
 
 def safety(optim_vars, aux_vars):
-    control = optim_vars[0].data.view(-1, 50, 2)
-    neighbors = aux_vars[0].data.permute(0, 2, 1, 3)
-    current_state = aux_vars[1].data
-    ref_line = aux_vars[2].data
+    control = optim_vars[0].tensor.view(-1, 50, 2)
+    neighbors = aux_vars[0].tensor.permute(0, 2, 1, 3)
+    current_state = aux_vars[1].tensor
+    ref_line = aux_vars[2].tensor
 
     actor_mask = torch.ne(current_state, 0)[:, 1:, -1]
     ego_current_state = current_state[:, 0]