Qustion about navhard two stage

[DanqiZhao21]

Hello team,

It seems that the NAVHard Two-Stage dataset zips are missing some LiDAR (MergedPointCloud) data, which prevents me from evaluating my end-to-end AD system. Could you please confirm whether I downloaded the data incorrectly, or if the dataset is indeed incomplete?

(wrapped_fn pid=3775615) File "/mnt/data/navsim_workspace/navsim/navsim/common/dataloader.py", line 257, in get_agent_input_from_token
(wrapped_fn pid=3775615) return Scene.load_from_disk(
(wrapped_fn pid=3775615) File "/mnt/data/navsim_workspace/navsim/navsim/common/dataclasses.py", line 699, in load_from_disk
(wrapped_fn pid=3775615) lidar = Lidar.from_paths(
(wrapped_fn pid=3775615) File "/mnt/data/navsim_workspace/navsim/navsim/common/dataclasses.py", line 132, in from_paths
(wrapped_fn pid=3775615) lidar_pc = LidarPointCloud.from_buffer(cls._load_bytes(global_lidar_path), "pcd").points
(wrapped_fn pid=3775615) File "/mnt/data/navsim_workspace/navsim/navsim/common/dataclasses.py", line 116, in _load_bytes
(wrapped_fn pid=3775615) with open(lidar_path, "rb") as fp:
(wrapped_fn pid=3775615) FileNotFoundError: [Errno 2] No such file or directory: '/mnt/data/navsim_workspace/dataset/navhard_two_stage/sensor_blobs/2021.09.16.19.27.01_veh-45_01749_03230/MergedPointCloud/ae0309beb1ab4cc70.pcd'
Ray objects: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 76/76 [06:38<00:00, 5.24s/it]
2025-10-06 16:37:04,507 WARNING {/mnt/data/navsim_workspace/navsim/navsim/planning/script/run_pdm_score.py:392} ----------- Failed to calculate pseudo closed-loop weights or comfort:
Traceback (most recent call last):

I compared the size of my downloaded data with the official dataset. Mine is 25 GB, whereas the official dataset is stated to be 33 GB.