[BugFix] fix the vehicle_capacity dimension mismatching bug

rl4co/envs/routing/cvrp.py

@@ -156,7 +156,7 @@ def _reset(
     def get_action_mask(td: TensorDict) -> torch.Tensor:
         # For demand steps_dim is inserted by indexing with id, for used_capacity insert node dim for broadcasting
         exceeds_cap = (
-            td["demand"][:, None, :] + td["used_capacity"][..., None] > td["vehicle_capacity"]
+            td["demand"][:, None, :] + td["used_capacity"][..., None] > td["vehicle_capacity"][..., None]
         )
 
         # Nodes that cannot be visited are already visited or too much demand to be served now