I do not want to train, but only want to test. I download pretrained "bpnet_5cm.pth.tar" and Set my config

DATA:
  data_name: scannet_cross
  data_root: /home/aaa/BPNet/ScanNet
  classes: 20
  aug: True
  voxelSize: 0.05



TRAIN:
  viewNum: 3
  weight_2d: 0.1
  arch: bpnet
  layers_2d: 34
  arch_3d: MinkUNet18A

  sync_bn_2d: True
  ignore_label: 255
  train_gpu: [0,1,2,3]
  workers: 16  # data loader workers
  batch_size: 16  # batch size for training
  batch_size_val: 16  # batch size for validation during training, memory and speed tradeoff
  base_lr: 0.01
  loop: 5
  epochs: 100
  start_epoch: 0
  power: 0.9
  momentum: 0.9
  weight_decay: 0.0001
  manual_seed: 1463
  print_freq: 10
  save_freq: 1
  save_path:
  weight:  # path to initial weight (default: none)
  resume:
  evaluate: True  # evaluate on validation set, extra gpu memory needed and small batch_size_val is recommend
  eval_freq: 1
#  zoom_factor: 8  # zoom factor for final prediction during training, be in [1, 2, 4, 8]
#  train_h: 241
#  train_w: 321
#  viewNum: 3

Distributed:
  dist_url: tcp://127.0.0.1:6787
  dist_backend: 'nccl'
  multiprocessing_distributed: True
  world_size: 1
  rank: 0


TEST:
  split: val  # split in [train, val and test]
  val_benchmark: True
  test_workers: 4
  test_gpu: [0]
  test_batch_size: 16
  model_path: /home/aaa/BPNet/Exp/scannet/EXP1/model/model_best.pth.tar
  save_folder: /home/aaa/BPNet/Exp/scannet/EXP1/result
  test_repeats: 7

And $ sh tool/test.sh EXP1 ./config/scannet/bpnet_5cm.yaml 8

And get

RuntimeError: Error(s) in loading state_dict for BPNet:
        Missing key(s) in state_dict: "layer0_2d.0.weight", "layer0_2d.1.weight", "layer0_2d.1.bias", "layer0_2d.1.running_mean", "layer0_2d.1.running_var", "layer1_2d.0.conv1.weight", "layer1_2d.0.bn1.weight", "layer1_2d.0.bn1.bias", "layer1_2d.0.bn1.running_mean", "layer1_2d.0.bn1.running_var", "layer1_2d.0.conv2.weight", "layer1_2d.0.bn2.weight", "layer1_2d.0.bn2.bias", "layer1_2d.0.bn2.running_mean", "layer1_2d.0.bn2.running_var", "layer1_2d.1.conv1.weight", "layer1_2d.1.bn1.weight", "layer1_2d.1.bn1.bias", "layer1_2d.1.bn1.running_mean", "layer1_2d.1.bn1.running_var", "layer1_2d.1.conv2.weight", "layer1_2d.1.bn2.weight", "layer1_2d.1.bn2.bias", "layer1_2d.1.bn2.running_mean", "layer1_2d.1.bn2.running_var", "layer1_2d.2.conv1.weight", "layer1_2d.2.bn1.weight", "layer1_2d.2.bn1.bias", "layer1_2d.2.bn1.running_mean", "layer1_2d.2.bn1.running_var", "layer1_2d.2.conv2.weight", "layer1_2d.2.bn2.weight", "layer1_2d.2.bn2.bias", "layer1_2d.2.bn2.running_mean", "layer1_2d.2.bn2.running_var", "layer2_2d.0.conv1.weight", "layer2_2d.0.bn1.weight", "layer2_2d.0.bn1.bias", "layer2_2d.0.bn1.running_mean", "layer2_2d.0.bn1.running_var", "layer2_2d.0.conv2.weight", "layer2_2d.0.bn2.weight", "layer2_2d.0.bn2.bias", "layer2_2d.0.bn2.running_mean", "layer2_2d.0.bn2.running_var", "layer2_2d.0.downsample.0.weight", "layer2_2d.0.downsample.1.weight", "layer2_2d.0.downsample.1.bias", "layer2_2d.0.downsample.1.running_mean", "layer2_2d.0.downsample.1.running_var", "layer2_2d.1.conv1.weight", "layer2_2d.1.bn1.weight", "layer2_2d.1.bn1.bias", "layer2_2d.1.bn1.running_mean", "layer2_2d.1.bn1.running_var", "layer2_2d.1.conv2.weight", "layer2_2d.1.bn2.weight", "layer2_2d.1.bn2.bias", "layer2_2d.1.bn2.running_mean", "layer2_2d.1.bn2.running_var", "layer2_2d.2.conv1.weight", "layer2_2d.2.bn1.weight", "layer2_2d.2.bn1.bias", "layer2_2d.2.bn1.running_mean", "layer2_2d.2.bn1.running_var", "layer2_2d.2.conv2.weight", "layer2_2d.2.bn2.weight", "layer2_2d.2.bn2.bias", "layer2_2d.2.bn2.running_mean", "layer2_2d.2.bn2.running_var", "layer2_2d.3.conv1.weight", "layer2_2d.3.bn1.weight", "layer2_2d.3.bn1.bias", "layer2_2d.3.bn1.running_mean", "layer2_2d.3.bn1.running_var", "layer2_2d.3.conv2.weight", "layer2_2d.3.bn2.weight", "layer2_2d.3.bn2.bias", "layer2_2d.3.bn2.running_mean", "layer2_2d.3.bn2.running_var", "layer3_2d.0.conv1.weight", "layer3_2d.0.bn1.weight", "layer3_2d.0.bn1.bias", "layer3_2d.0.bn1.running_mean", "layer3_2d.0.bn1.running_var", "layer3_2d.0.conv2.weight", "layer3_2d.0.bn2.weight", "layer3_2d.0.bn2.bias", "layer3_2d.0.bn2.running_mean", "layer3_2d.0.bn2.running_var", "layer3_2d.0.downsample.0.weight", "layer3_2d.0.downsample.1.weight", "layer3_2d.0.downsample.1.bias", "layer3_2d.0.downsample.1.running_mean", "layer3_2d.0.downsample.1.running_var", "layer3_2d.1.conv1.weight", "layer3_2d.1.bn1.weight", "layer3_2d.1.bn1.bias", "layer3_2d.1.bn1.running_mean", "layer3_2d.1.bn1.running_var", "layer3_2d.1.conv2.weight", "layer3_2d.1.bn2.weight", "layer3_2d.1.bn2.bias", "layer3_2d.1.bn2.running_mean", "layer3_2d.1.bn2.running_var", "layer3_2d.2.conv1.weight", "layer3_2d.2.bn1.weight", "layer3_2d.2.bn1.bias", "layer3_2d.2.bn1.running_mean", "layer3_2d.2.bn1.running_var", "layer3_2d.2.conv2.weight", "layer3_2d.2.bn2.weight", "layer3_2d.2.bn2.bias", "layer3_2d.2.bn2.running_mean", "layer3_2d.2.bn2.running_var", "layer3_2d.3.conv1.weight", "layer3_2d.3.bn1.weight", "layer3_2d.3.bn1.bias", "layer3_2d.3.bn1.running_mean", "layer3_2d.3.bn1.running_var", "layer3_2d.3.conv2.weight", "layer3_2d.3.bn2.weight", "layer3_2d.3.bn2.bias", "layer3_2d.3.bn2.running_mean", "layer3_2d.3.bn2.running_var", "layer3_2d.4.conv1.weight", "layer3_2d.4.bn1.weight", "layer3_2d.4.bn1.bias", "layer3_2d.4.bn1.running_mean", "layer3_2d.4.bn1.running_var", "layer3_2d.4.conv2.weight", "layer3_2d.4.bn2.weight", "layer3_2d.4.bn2.bias", "layer3_2d.4.bn2.running_mean", "layer3_2d.4.bn2.running_var", "layer3_2d.5.conv1.weight", "layer3_2d.5.bn1.weight", "layer3_2d.5.bn1.bias", "layer3_2d.5.bn1.running_mean", "layer3_2d.5.bn1.running_var", "layer3_2d.5.conv2.weight", "layer3_2d.5.bn2.weight", "layer3_2d.5.bn2.bias", "layer3_2d.5.bn2.running_mean", "layer3_2d.5.bn2.running_var", "layer4_2d.0.conv1.weight", "layer4_2d.0.bn1.weight", "layer4_2d.0.bn1.bias", "layer4_2d.0.bn1.running_mean", "layer4_2d.0.bn1.running_var", "layer4_2d.0.conv2.weight", "layer4_2d.0.bn2.weight", "layer4_2d.0.bn2.bias", "layer4_2d.0.bn2.running_mean", "layer4_2d.0.bn2.running_var", "layer4_2d.0.downsample.0.weight", "layer4_2d.0.downsample.1.weight", "layer4_2d.0.downsample.1.bias", "layer4_2d.0.downsample.1.running_mean", "layer4_2d.0.downsample.1.running_var", "layer4_2d.1.conv1.weight", "layer4_2d.1.bn1.weight", "layer4_2d.1.bn1.bias", "layer4_2d.1.bn1.running_mean", "layer4_2d.1.bn1.running_var", "layer4_2d.1.conv2.weight", "layer4_2d.1.bn2.weight", "layer4_2d.1.bn2.bias", "layer4_2d.1.bn2.running_mean", "layer4_2d.1.bn2.running_var", "layer4_2d.2.conv1.weight", "layer4_2d.2.bn1.weight", "layer4_2d.2.bn1.bias", "layer4_2d.2.bn1.running_mean", "layer4_2d.2.bn1.running_var", "layer4_2d.2.conv2.weight", "layer4_2d.2.bn2.weight", "layer4_2d.2.bn2.bias", "layer4_2d.2.bn2.running_mean", "layer4_2d.2.bn2.running_var", "up4_2d.0.weight", "up4_2d.0.bias", "up4_2d.1.weight", "up4_2d.1.bias", "up4_2d.1.running_mean", "up4_2d.1.running_var", "delayer4_2d.0.conv1.weight", "delayer4_2d.0.bn1.weight", "delayer4_2d.0.bn1.bias", "delayer4_2d.0.bn1.running_mean", "delayer4_2d.0.bn1.running_var", "delayer4_2d.0.conv2.weight", "delayer4_2d.0.bn2.weight", "delayer4_2d.0.bn2.bias", "delayer4_2d.0.bn2.running_mean", "delayer4_2d.0.bn2.running_var", "delayer4_2d.0.downsample.0.weight", "delayer4_2d.0.downsample.1.weight", "delayer4_2d.0.downsample.1.bias", "delayer4_2d.0.downsample.1.running_mean", "delayer4_2d.0.downsample.1.running_var", "delayer4_2d.1.conv1.weight", "delayer4_2d.1.bn1.weight", "delayer4_2d.1.bn1.bias", "delayer4_2d.1.bn1.running_mean", "delayer4_2d.1.bn1.running_var", "delayer4_2d.1.conv2.weight", "delayer4_2d.1.bn2.weight", "delayer4_2d.1.bn2.bias", "delayer4_2d.1.bn2.running_mean", "delayer4_2d.1.bn2.running_var", "delayer4_2d.2.conv1.weight", "delayer4_2d.2.bn1.weight", "delayer4_2d.2.bn1.bias", "delayer4_2d.2.bn1.running_mean", "delayer4_2d.2.bn1.running_var", "delayer4_2d.2.conv2.weight", "delayer4_2d.2.bn2.weight", "delayer4_2d.2.bn2.bias", "delayer4_2d.2.bn2.running_mean", "delayer4_2d.2.bn2.running_var", "up3_2d.0.weight", "up3_2d.0.bias", "up3_2d.1.weight", "up3_2d.1.bias", "up3_2d.1.running_mean", "up3_2d.1.running_var", "delayer3_2d.0.conv1.weight", "delayer3_2d.0.bn1.weight", "delayer3_2d.0.bn1.bias", "delayer3_2d.0.bn1.running_mean", "delayer3_2d.0.bn1.running_var", "delayer3_2d.0.conv2.weight", "delayer3_2d.0.bn2.weight", "delayer3_2d.0.bn2.bias", "delayer3_2d.0.bn2.running_mean", "delayer3_2d.0.bn2.running_var", "delayer3_2d.0.downsample.0.weight", "delayer3_2d.0.downsample.1.weight", "delayer3_2d.0.downsample.1.bias", "delayer3_2d.0.downsample.1.running_mean", "delayer3_2d.0.downsample.1.running_var", "delayer3_2d.1.conv1.weight", "delayer3_2d.1.bn1.weight", "delayer3_2d.1.bn1.bias", "delayer3_2d.1.bn1.running_mean", "delayer3_2d.1.bn1.running_var", "delayer3_2d.1.conv2.weight", "delayer3_2d.1.bn2.weight", "delayer3_2d.1.bn2.bias", "delayer3_2d.1.bn2.running_mean", "delayer3_2d.1.bn2.running_var", "delayer3_2d.2.conv1.weight", "delayer3_2d.2.bn1.weight", "delayer3_2d.2.bn1.bias", "delayer3_2d.2.bn1.running_mean", "delayer3_2d.2.bn1.running_var", "delayer3_2d.2.conv2.weight", "delayer3_2d.2.bn2.weight", "delayer3_2d.2.bn2.bias", "delayer3_2d.2.bn2.running_mean", "delayer3_2d.2.bn2.running_var", "delayer3_2d.3.conv1.weight", "delayer3_2d.3.bn1.weight", "delayer3_2d.3.bn1.bias", "delayer3_2d.3.bn1.running_mean", "delayer3_2d.3.bn1.running_var", "delayer3_2d.3.conv2.weight", "delayer3_2d.3.bn2.weight", "delayer3_2d.3.bn2.bias", "delayer3_2d.3.bn2.running_mean", "delayer3_2d.3.bn2.running_var", "delayer3_2d.4.conv1.weight", "delayer3_2d.4.bn1.weight", "delayer3_2d.4.bn1.bias", "delayer3_2d.4.bn1.running_mean", "delayer3_2d.4.bn1.running_var", "delayer3_2d.4.conv2.weight", "delayer3_2d.4.bn2.weight", "delayer3_2d.4.bn2.bias", "delayer3_2d.4.bn2.running_mean", "delayer3_2d.4.bn2.running_var", "delayer3_2d.5.conv1.weight", "delayer3_2d.5.bn1.weight", "delayer3_2d.5.bn1.bias", "delayer3_2d.5.bn1.running_mean", "delayer3_2d.5.bn1.running_var", "delayer3_2d.5.conv2.weight", "delayer3_2d.5.bn2.weight", "delayer3_2d.5.bn2.bias", "delayer3_2d.5.bn2.running_mean", "delayer3_2d.5.bn2.running_var", "up2_2d.0.weight", "up2_2d.0.bias", "up2_2d.1.weight", "up2_2d.1.bias", "up2_2d.1.running_mean", "up2_2d.1.running_var", "delayer2_2d.0.conv1.weight", "delayer2_2d.0.bn1.weight", "delayer2_2d.0.bn1.bias", "delayer2_2d.0.bn1.running_mean", "delayer2_2d.0.bn1.running_var", "delayer2_2d.0.conv2.weight", "delayer2_2d.0.bn2.weight", "delayer2_2d.0.bn2.bias", "delayer2_2d.0.bn2.running_mean", "delayer2_2d.0.bn2.running_var", "delayer2_2d.0.downsample.0.weight", "delayer2_2d.0.downsample.1.weight", "delayer2_2d.0.downsample.1.bias", "delayer2_2d.0.downsample.1.running_mean", "delayer2_2d.0.downsample.1.running_var", "delayer2_2d.1.conv1.weight", "delayer2_2d.1.bn1.weight", "delayer2_2d.1.bn1.bias", "delayer2_2d.1.bn1.running_mean", "delayer2_2d.1.bn1.running_var", "delayer2_2d.1.conv2.weight", "delayer2_2d.1.bn2.weight", "delayer2_2d.1.bn2.bias", "delayer2_2d.1.bn2.running_mean", "delayer2_2d.1.bn2.running_var", "delayer2_2d.2.conv1.weight", "delayer2_2d.2.bn1.weight", "delayer2_2d.2.bn1.bias", "delayer2_2d.2.bn1.running_mean", "delayer2_2d.2.bn1.running_var", "delayer2_2d.2.conv2.weight", "delayer2_2d.2.bn2.weight", "delayer2_2d.2.bn2.bias", "delayer2_2d.2.bn2.running_mean", "delayer2_2d.2.bn2.running_var", "delayer2_2d.3.conv1.weight", "delayer2_2d.3.bn1.weight", "delayer2_2d.3.bn1.bias", "delayer2_2d.3.bn1.running_mean", "delayer2_2d.3.bn1.running_var", "delayer2_2d.3.conv2.weight", "delayer2_2d.3.bn2.weight", "delayer2_2d.3.bn2.bias", "delayer2_2d.3.bn2.running_mean", "delayer2_2d.3.bn2.running_var", "cls_2d.0.weight", "cls_2d.1.weight", "cls_2d.1.bias", "cls_2d.1.running_mean", "cls_2d.1.running_var", "cls_2d.3.weight", "cls_2d.3.bias", "layer0_3d.0.kernel", "layer0_3d.1.bn.weight", "layer0_3d.1.bn.bias", "layer0_3d.1.bn.running_mean", "layer0_3d.1.bn.running_var", "layer1_3d.0.kernel", "layer1_3d.1.bn.weight", "layer1_3d.1.bn.bias", "layer1_3d.1.bn.running_mean", "layer1_3d.1.bn.running_var", "layer1_3d.3.0.conv1.kernel", "layer1_3d.3.0.norm1.bn.weight", "layer1_3d.3.0.norm1.bn.bias", "layer1_3d.3.0.norm1.bn.running_mean", "layer1_3d.3.0.norm1.bn.running_var", "layer1_3d.3.0.conv2.kernel", "layer1_3d.3.0.norm2.bn.weight", "layer1_3d.3.0.norm2.bn.bias", "layer1_3d.3.0.norm2.bn.running_mean", "layer1_3d.3.0.norm2.bn.running_var", "layer1_3d.3.1.conv1.kernel", "layer1_3d.3.1.norm1.bn.weight", "layer1_3d.3.1.norm1.bn.bias", "layer1_3d.3.1.norm1.bn.running_mean", "layer1_3d.3.1.norm1.bn.running_var", "layer1_3d.3.1.conv2.kernel", "layer1_3d.3.1.norm2.bn.weight", "layer1_3d.3.1.norm2.bn.bias", "layer1_3d.3.1.norm2.bn.running_mean", "layer1_3d.3.1.norm2.bn.running_var", "layer2_3d.0.kernel", "layer2_3d.1.bn.weight", "layer2_3d.1.bn.bias", "layer2_3d.1.bn.running_mean", "layer2_3d.1.bn.running_var", "layer2_3d.3.0.conv1.kernel", "layer2_3d.3.0.norm1.bn.weight", "layer2_3d.3.0.norm1.bn.bias", "layer2_3d.3.0.norm1.bn.running_mean", "layer2_3d.3.0.norm1.bn.running_var", "layer2_3d.3.0.conv2.kernel", "layer2_3d.3.0.norm2.bn.weight", "layer2_3d.3.0.norm2.bn.bias", "layer2_3d.3.0.norm2.bn.running_mean", "layer2_3d.3.0.norm2.bn.running_var", "layer2_3d.3.0.downsample.0.kernel", "layer2_3d.3.0.downsample.1.bn.weight", "layer2_3d.3.0.downsample.1.bn.bias", "layer2_3d.3.0.downsample.1.bn.running_mean", "layer2_3d.3.0.downsample.1.bn.running_var", "layer2_3d.3.1.conv1.kernel", "layer2_3d.3.1.norm1.bn.weight", "layer2_3d.3.1.norm1.bn.bias", "layer2_3d.3.1.norm1.bn.running_mean", "layer2_3d.3.1.norm1.bn.running_var", "layer2_3d.3.1.conv2.kernel", "layer2_3d.3.1.norm2.bn.weight", "layer2_3d.3.1.norm2.bn.bias", "layer2_3d.3.1.norm2.bn.running_mean", "layer2_3d.3.1.norm2.bn.running_var", "layer3_3d.0.kernel", "layer3_3d.1.bn.weight", "layer3_3d.1.bn.bias", "layer3_3d.1.bn.running_mean", "layer3_3d.1.bn.running_var", "layer3_3d.3.0.conv1.kernel", "layer3_3d.3.0.norm1.bn.weight", "layer3_3d.3.0.norm1.bn.bias", "layer3_3d.3.0.norm1.bn.running_mean", "layer3_3d.3.0.norm1.bn.running_var", "layer3_3d.3.0.conv2.kernel", "layer3_3d.3.0.norm2.bn.weight", "layer3_3d.3.0.norm2.bn.bias", "layer3_3d.3.0.norm2.bn.running_mean", "layer3_3d.3.0.norm2.bn.running_var", "layer3_3d.3.0.downsample.0.kernel", "layer3_3d.3.0.downsample.1.bn.weight", "layer3_3d.3.0.downsample.1.bn.bias", "layer3_3d.3.0.downsample.1.bn.running_mean", "layer3_3d.3.0.downsample.1.bn.running_var", "layer3_3d.3.1.conv1.kernel", "layer3_3d.3.1.norm1.bn.weight", "layer3_3d.3.1.norm1.bn.bias", "layer3_3d.3.1.norm1.bn.running_mean", "layer3_3d.3.1.norm1.bn.running_var", "layer3_3d.3.1.conv2.kernel", "layer3_3d.3.1.norm2.bn.weight", "layer3_3d.3.1.norm2.bn.bias", "layer3_3d.3.1.norm2.bn.running_mean", "layer3_3d.3.1.norm2.bn.running_var", "layer4_3d.0.kernel", "layer4_3d.1.bn.weight", "layer4_3d.1.bn.bias", "layer4_3d.1.bn.running_mean", "layer4_3d.1.bn.running_var", "layer4_3d.3.0.conv1.kernel", "layer4_3d.3.0.norm1.bn.weight", "layer4_3d.3.0.norm1.bn.bias", "layer4_3d.3.0.norm1.bn.running_mean", "layer4_3d.3.0.norm1.bn.running_var", "layer4_3d.3.0.conv2.kernel", "layer4_3d.3.0.norm2.bn.weight", "layer4_3d.3.0.norm2.bn.bias", "layer4_3d.3.0.norm2.bn.running_mean", "layer4_3d.3.0.norm2.bn.running_var", "layer4_3d.3.0.downsample.0.kernel", "layer4_3d.3.0.downsample.1.bn.weight", "layer4_3d.3.0.downsample.1.bn.bias", "layer4_3d.3.0.downsample.1.bn.running_mean", "layer4_3d.3.0.downsample.1.bn.running_var", "layer4_3d.3.1.conv1.kernel", "layer4_3d.3.1.norm1.bn.weight", "layer4_3d.3.1.norm1.bn.bias", "layer4_3d.3.1.norm1.bn.running_mean", "layer4_3d.3.1.norm1.bn.running_var", "layer4_3d.3.1.conv2.kernel", "layer4_3d.3.1.norm2.bn.weight", "layer4_3d.3.1.norm2.bn.bias", "layer4_3d.3.1.norm2.bn.running_mean", "layer4_3d.3.1.norm2.bn.running_var", "layer5_3d.0.kernel", "layer5_3d.1.bn.weight", "layer5_3d.1.bn.bias", "layer5_3d.1.bn.running_mean", "layer5_3d.1.bn.running_var", "layer6_3d.0.0.conv1.kernel", "layer6_3d.0.0.norm1.bn.weight", "layer6_3d.0.0.norm1.bn.bias", "layer6_3d.0.0.norm1.bn.running_mean", "layer6_3d.0.0.norm1.bn.running_var", "layer6_3d.0.0.conv2.kernel", "layer6_3d.0.0.norm2.bn.weight", "layer6_3d.0.0.norm2.bn.bias", "layer6_3d.0.0.norm2.bn.running_mean", "layer6_3d.0.0.norm2.bn.running_var", "layer6_3d.0.0.downsample.0.kernel", "layer6_3d.0.0.downsample.1.bn.weight", "layer6_3d.0.0.downsample.1.bn.bias", "layer6_3d.0.0.downsample.1.bn.running_mean", "layer6_3d.0.0.downsample.1.bn.running_var", "layer6_3d.0.1.conv1.kernel", "layer6_3d.0.1.norm1.bn.weight", "layer6_3d.0.1.norm1.bn.bias", "layer6_3d.0.1.norm1.bn.running_mean", "layer6_3d.0.1.norm1.bn.running_var", "layer6_3d.0.1.conv2.kernel", "layer6_3d.0.1.norm2.bn.weight", "layer6_3d.0.1.norm2.bn.bias", "layer6_3d.0.1.norm2.bn.running_mean", "layer6_3d.0.1.norm2.bn.running_var", "layer6_3d.1.kernel", "layer6_3d.2.bn.weight", "layer6_3d.2.bn.bias", "layer6_3d.2.bn.running_mean", "layer6_3d.2.bn.running_var", "layer7_3d.0.0.conv1.kernel", "layer7_3d.0.0.norm1.bn.weight", "layer7_3d.0.0.norm1.bn.bias", "layer7_3d.0.0.norm1.bn.running_mean", "layer7_3d.0.0.norm1.bn.running_var", "layer7_3d.0.0.conv2.kernel", "layer7_3d.0.0.norm2.bn.weight", "layer7_3d.0.0.norm2.bn.bias", "layer7_3d.0.0.norm2.bn.running_mean", "layer7_3d.0.0.norm2.bn.running_var", "layer7_3d.0.0.downsample.0.kernel", "layer7_3d.0.0.downsample.1.bn.weight", "layer7_3d.0.0.downsample.1.bn.bias", "layer7_3d.0.0.downsample.1.bn.running_mean", "layer7_3d.0.0.downsample.1.bn.running_var", "layer7_3d.0.1.conv1.kernel", "layer7_3d.0.1.norm1.bn.weight", "layer7_3d.0.1.norm1.bn.bias", "layer7_3d.0.1.norm1.bn.running_mean", "layer7_3d.0.1.norm1.bn.running_var", "layer7_3d.0.1.conv2.kernel", "layer7_3d.0.1.norm2.bn.weight", "layer7_3d.0.1.norm2.bn.bias", "layer7_3d.0.1.norm2.bn.running_mean", "layer7_3d.0.1.norm2.bn.running_var", "layer7_3d.1.kernel", "layer7_3d.2.bn.weight", "layer7_3d.2.bn.bias", "layer7_3d.2.bn.running_mean", "layer7_3d.2.bn.running_var", "layer8_3d.0.0.conv1.kernel", "layer8_3d.0.0.norm1.bn.weight", "layer8_3d.0.0.norm1.bn.bias", "layer8_3d.0.0.norm1.bn.running_mean", "layer8_3d.0.0.norm1.bn.running_var", "layer8_3d.0.0.conv2.kernel", "layer8_3d.0.0.norm2.bn.weight", "layer8_3d.0.0.norm2.bn.bias", "layer8_3d.0.0.norm2.bn.running_mean", "layer8_3d.0.0.norm2.bn.running_var", "layer8_3d.0.0.downsample.0.kernel", "layer8_3d.0.0.downsample.1.bn.weight", "layer8_3d.0.0.downsample.1.bn.bias", "layer8_3d.0.0.downsample.1.bn.running_mean", "layer8_3d.0.0.downsample.1.bn.running_var", "layer8_3d.0.1.conv1.kernel", "layer8_3d.0.1.norm1.bn.weight", "layer8_3d.0.1.norm1.bn.bias", "layer8_3d.0.1.norm1.bn.running_mean", "layer8_3d.0.1.norm1.bn.running_var", "layer8_3d.0.1.conv2.kernel", "layer8_3d.0.1.norm2.bn.weight", "layer8_3d.0.1.norm2.bn.bias", "layer8_3d.0.1.norm2.bn.running_mean", "layer8_3d.0.1.norm2.bn.running_var", "layer8_3d.1.kernel", "layer8_3d.2.bn.weight", "layer8_3d.2.bn.bias", "layer8_3d.2.bn.running_mean", "layer8_3d.2.bn.running_var", "layer9_3d.0.conv1.kernel", "layer9_3d.0.norm1.bn.weight", "layer9_3d.0.norm1.bn.bias", "layer9_3d.0.norm1.bn.running_mean", "layer9_3d.0.norm1.bn.running_var", "layer9_3d.0.conv2.kernel", "layer9_3d.0.norm2.bn.weight", "layer9_3d.0.norm2.bn.bias", "layer9_3d.0.norm2.bn.running_mean", "layer9_3d.0.norm2.bn.running_var", "layer9_3d.0.downsample.0.kernel", "layer9_3d.0.downsample.1.bn.weight", "layer9_3d.0.downsample.1.bn.bias", "layer9_3d.0.downsample.1.bn.running_mean", "layer9_3d.0.downsample.1.bn.running_var", "layer9_3d.1.conv1.kernel", "layer9_3d.1.norm1.bn.weight", "layer9_3d.1.norm1.bn.bias", "layer9_3d.1.norm1.bn.running_mean", "layer9_3d.1.norm1.bn.running_var", "layer9_3d.1.conv2.kernel", "layer9_3d.1.norm2.bn.weight", "layer9_3d.1.norm2.bn.bias", "layer9_3d.1.norm2.bn.running_mean", "layer9_3d.1.norm2.bn.running_var", "cls_3d.kernel", "cls_3d.bias", "linker_p2.view_fusion.0.kernel", "linker_p2.view_fusion.1.bn.weight", "linker_p2.view_fusion.1.bn.bias", "linker_p2.view_fusion.1.bn.running_mean", "linker_p2.view_fusion.1.bn.running_var", "linker_p2.view_fusion.3.kernel", "linker_p2.view_fusion.4.bn.weight", "linker_p2.view_fusion.4.bn.bias", "linker_p2.view_fusion.4.bn.running_mean", "linker_p2.view_fusion.4.bn.running_var", "linker_p2.fuseTo3d.0.kernel", "linker_p2.fuseTo3d.1.bn.weight", "linker_p2.fuseTo3d.1.bn.bias", "linker_p2.fuseTo3d.1.bn.running_mean", "linker_p2.fuseTo3d.1.bn.running_var", "linker_p2.view_sep.0.kernel", "linker_p2.view_sep.1.bn.weight", "linker_p2.view_sep.1.bn.bias", "linker_p2.view_sep.1.bn.running_mean", "linker_p2.view_sep.1.bn.running_var", "linker_p2.fuseTo2d.0.weight", "linker_p2.fuseTo2d.1.weight", "linker_p2.fuseTo2d.1.bias", "linker_p2.fuseTo2d.1.running_mean", "linker_p2.fuseTo2d.1.running_var", "linker_p3.view_fusion.0.kernel", "linker_p3.view_fusion.1.bn.weight", "linker_p3.view_fusion.1.bn.bias", "linker_p3.view_fusion.1.bn.running_mean", "linker_p3.view_fusion.1.bn.running_var", "linker_p3.view_fusion.3.kernel", "linker_p3.view_fusion.4.bn.weight", "linker_p3.view_fusion.4.bn.bias", "linker_p3.view_fusion.4.bn.running_mean", "linker_p3.view_fusion.4.bn.running_var", "linker_p3.fuseTo3d.0.kernel", "linker_p3.fuseTo3d.1.bn.weight", "linker_p3.fuseTo3d.1.bn.bias", "linker_p3.fuseTo3d.1.bn.running_mean", "linker_p3.fuseTo3d.1.bn.running_var", "linker_p3.view_sep.0.kernel", "linker_p3.view_sep.1.bn.weight", "linker_p3.view_sep.1.bn.bias", "linker_p3.view_sep.1.bn.running_mean", "linker_p3.view_sep.1.bn.running_var", "linker_p3.fuseTo2d.0.weight", "linker_p3.fuseTo2d.1.weight", "linker_p3.fuseTo2d.1.bias", "linker_p3.fuseTo2d.1.running_mean", "linker_p3.fuseTo2d.1.running_var", "linker_p4.view_fusion.0.kernel", "linker_p4.view_fusion.1.bn.weight", "linker_p4.view_fusion.1.bn.bias", "linker_p4.view_fusion.1.bn.running_mean", "linker_p4.view_fusion.1.bn.running_var", "linker_p4.view_fusion.3.kernel", "linker_p4.view_fusion.4.bn.weight", "linker_p4.view_fusion.4.bn.bias", "linker_p4.view_fusion.4.bn.running_mean", "linker_p4.view_fusion.4.bn.running_var", "linker_p4.fuseTo3d.0.kernel", "linker_p4.fuseTo3d.1.bn.weight", "linker_p4.fuseTo3d.1.bn.bias", "linker_p4.fuseTo3d.1.bn.running_mean", "linker_p4.fuseTo3d.1.bn.running_var", "linker_p4.view_sep.0.kernel", "linker_p4.view_sep.1.bn.weight", "linker_p4.view_sep.1.bn.bias", "linker_p4.view_sep.1.bn.running_mean", "linker_p4.view_sep.1.bn.running_var", "linker_p4.fuseTo2d.0.weight", "linker_p4.fuseTo2d.1.weight", "linker_p4.fuseTo2d.1.bias", "linker_p4.fuseTo2d.1.running_mean", "linker_p4.fuseTo2d.1.running_var", "linker_p5.view_fusion.0.kernel", "linker_p5.view_fusion.1.bn.weight", "linker_p5.view_fusion.1.bn.bias", "linker_p5.view_fusion.1.bn.running_mean", "linker_p5.view_fusion.1.bn.running_var", "linker_p5.view_fusion.3.kernel", "linker_p5.view_fusion.4.bn.weight", "linker_p5.view_fusion.4.bn.bias", "linker_p5.view_fusion.4.bn.running_mean", "linker_p5.view_fusion.4.bn.running_var", "linker_p5.fuseTo3d.0.kernel", "linker_p5.fuseTo3d.1.bn.weight", "linker_p5.fuseTo3d.1.bn.bias", "linker_p5.fuseTo3d.1.bn.running_mean", "linker_p5.fuseTo3d.1.bn.running_var", "linker_p5.view_sep.0.kernel", "linker_p5.view_sep.1.bn.weight", "linker_p5.view_sep.1.bn.bias", "linker_p5.view_sep.1.bn.running_mean", "linker_p5.view_sep.1.bn.running_var", "linker_p5.fuseTo2d.0.weight", "linker_p5.fuseTo2d.1.weight", "linker_p5.fuseTo2d.1.bias", "linker_p5.fuseTo2d.1.running_mean", "linker_p5.fuseTo2d.1.running_var". 
        Unexpected key(s) in state_dict: "module.layer0_2d.0.weight", "module.layer0_2d.1.weight", "module.layer0_2d.1.bias", "module.layer0_2d.1.running_mean", "module.layer0_2d.1.running_var", "module.layer0_2d.1.num_batches_tracked", "module.layer1_2d.0.conv1.weight", "module.layer1_2d.0.bn1.weight", "module.layer1_2d.0.bn1.bias", "module.layer1_2d.0.bn1.running_mean", "module.layer1_2d.0.bn1.running_var", "module.layer1_2d.0.bn1.num_batches_tracked", "module.layer1_2d.0.conv2.weight", "module.layer1_2d.0.bn2.weight", "module.layer1_2d.0.bn2.bias", "module.layer1_2d.0.bn2.running_mean", "module.layer1_2d.0.bn2.running_var", "module.layer1_2d.0.bn2.num_batches_tracked", "module.layer1_2d.1.conv1.weight", "module.layer1_2d.1.bn1.weight", "module.layer1_2d.1.bn1.bias", "module.layer1_2d.1.bn1.running_mean", "module.layer1_2d.1.bn1.running_var", "module.layer1_2d.1.bn1.num_batches_tracked", "module.layer1_2d.1.conv2.weight", "module.layer1_2d.1.bn2.weight", "module.layer1_2d.1.bn2.bias", "module.layer1_2d.1.bn2.running_mean", "module.layer1_2d.1.bn2.running_var", "module.layer1_2d.1.bn2.num_batches_tracked", "module.layer1_2d.2.conv1.weight", "module.layer1_2d.2.bn1.weight", "module.layer1_2d.2.bn1.bias", "module.layer1_2d.2.bn1.running_mean", "module.layer1_2d.2.bn1.running_var", "module.layer1_2d.2.bn1.num_batches_tracked", "module.layer1_2d.2.conv2.weight", "module.layer1_2d.2.bn2.weight", "module.layer1_2d.2.bn2.bias", "module.layer1_2d.2.bn2.running_mean", "module.layer1_2d.2.bn2.running_var", "module.layer1_2d.2.bn2.num_batches_tracked", "module.layer2_2d.0.conv1.weight", "module.layer2_2d.0.bn1.weight", "module.layer2_2d.0.bn1.bias", "module.layer2_2d.0.bn1.running_mean", "module.layer2_2d.0.bn1.running_var", "module.layer2_2d.0.bn1.num_batches_tracked", "module.layer2_2d.0.conv2.weight", "module.layer2_2d.0.bn2.weight", "module.layer2_2d.0.bn2.bias", "module.layer2_2d.0.bn2.running_mean", "module.layer2_2d.0.bn2.running_var", "module.layer2_2d.0.bn2.num_batches_tracked", "module.layer2_2d.0.downsample.0.weight", "module.layer2_2d.0.downsample.1.weight", "module.layer2_2d.0.downsample.1.bias", "module.layer2_2d.0.downsample.1.running_mean", "module.layer2_2d.0.downsample.1.running_var", "module.layer2_2d.0.downsample.1.num_batches_tracked", "module.layer2_2d.1.conv1.weight", "module.layer2_2d.1.bn1.weight", "module.layer2_2d.1.bn1.bias", "module.layer2_2d.1.bn1.running_mean", "module.layer2_2d.1.bn1.running_var", "module.layer2_2d.1.bn1.num_batches_tracked", "module.layer2_2d.1.conv2.weight", "module.layer2_2d.1.bn2.weight", "module.layer2_2d.1.bn2.bias", "module.layer2_2d.1.bn2.running_mean", "module.layer2_2d.1.bn2.running_var", "module.layer2_2d.1.bn2.num_batches_tracked", "module.layer2_2d.2.conv1.weight", "module.layer2_2d.2.bn1.weight", "module.layer2_2d.2.bn1.bias", "module.layer2_2d.2.bn1.running_mean", "module.layer2_2d.2.bn1.running_var", "module.layer2_2d.2.bn1.num_batches_tracked", "module.layer2_2d.2.conv2.weight", "module.layer2_2d.2.bn2.weight", "module.layer2_2d.2.bn2.bias", "module.layer2_2d.2.bn2.running_mean", "module.layer2_2d.2.bn2.running_var", "module.layer2_2d.2.bn2.num_batches_tracked", "module.layer2_2d.3.conv1.weight", "module.layer2_2d.3.bn1.weight", "module.layer2_2d.3.bn1.bias", "module.layer2_2d.3.bn1.running_mean", "module.layer2_2d.3.bn1.running_var", "module.layer2_2d.3.bn1.num_batches_tracked", "module.layer2_2d.3.conv2.weight", "module.layer2_2d.3.bn2.weight", "module.layer2_2d.3.bn2.bias", "module.layer2_2d.3.bn2.running_mean", "module.layer2_2d.3.bn2.running_var", "module.layer2_2d.3.bn2.num_batches_tracked", "module.layer3_2d.0.conv1.weight", "module.layer3_2d.0.bn1.weight", "module.layer3_2d.0.bn1.bias", "module.layer3_2d.0.bn1.running_mean", "module.layer3_2d.0.bn1.running_var", "module.layer3_2d.0.bn1.num_batches_tracked", "module.layer3_2d.0.conv2.weight", "module.layer3_2d.0.bn2.weight", "module.layer3_2d.0.bn2.bias", "module.layer3_2d.0.bn2.running_mean", "module.layer3_2d.0.bn2.running_var", "module.layer3_2d.0.bn2.num_batches_tracked", "module.layer3_2d.0.downsample.0.weight", "module.layer3_2d.0.downsample.1.weight", "module.layer3_2d.0.downsample.1.bias", "module.layer3_2d.0.downsample.1.running_mean", "module.layer3_2d.0.downsample.1.running_var", "module.layer3_2d.0.downsample.1.num_batches_tracked", "module.layer3_2d.1.conv1.weight", "module.layer3_2d.1.bn1.weight", "module.layer3_2d.1.bn1.bias", "module.layer3_2d.1.bn1.running_mean", "module.layer3_2d.1.bn1.running_var", "module.layer3_2d.1.bn1.num_batches_tracked", "module.layer3_2d.1.conv2.weight", "module.layer3_2d.1.bn2.weight", "module.layer3_2d.1.bn2.bias", "module.layer3_2d.1.bn2.running_mean", "module.layer3_2d.1.bn2.running_var", "module.layer3_2d.1.bn2.num_batches_tracked", "module.layer3_2d.2.conv1.weight", "module.layer3_2d.2.bn1.weight", "module.layer3_2d.2.bn1.bias", "module.layer3_2d.2.bn1.running_mean", "module.layer3_2d.2.bn1.running_var", "module.layer3_2d.2.bn1.num_batches_tracked", "module.layer3_2d.2.conv2.weight", "module.layer3_2d.2.bn2.weight", "module.layer3_2d.2.bn2.bias", "module.layer3_2d.2.bn2.running_mean", "module.layer3_2d.2.bn2.running_var", "module.layer3_2d.2.bn2.num_batches_tracked", "module.layer3_2d.3.conv1.weight", "module.layer3_2d.3.bn1.weight", "module.layer3_2d.3.bn1.bias", "module.layer3_2d.3.bn1.running_mean", "module.layer3_2d.3.bn1.running_var", "module.layer3_2d.3.bn1.num_batches_tracked", "module.layer3_2d.3.conv2.weight", "module.layer3_2d.3.bn2.weight", "module.layer3_2d.3.bn2.bias", "module.layer3_2d.3.bn2.running_mean", "module.layer3_2d.3.bn2.running_var", "module.layer3_2d.3.bn2.num_batches_tracked", "module.layer3_2d.4.conv1.weight", "module.layer3_2d.4.bn1.weight", "module.layer3_2d.4.bn1.bias", "module.layer3_2d.4.bn1.running_mean", "module.layer3_2d.4.bn1.running_var", "module.layer3_2d.4.bn1.num_batches_tracked", "module.layer3_2d.4.conv2.weight", "module.layer3_2d.4.bn2.weight", "module.layer3_2d.4.bn2.bias", "module.layer3_2d.4.bn2.running_mean", "module.layer3_2d.4.bn2.running_var", "module.layer3_2d.4.bn2.num_batches_tracked", "module.layer3_2d.5.conv1.weight", "module.layer3_2d.5.bn1.weight", "module.layer3_2d.5.bn1.bias", "module.layer3_2d.5.bn1.running_mean", "module.layer3_2d.5.bn1.running_var", "module.layer3_2d.5.bn1.num_batches_tracked", "module.layer3_2d.5.conv2.weight", "module.layer3_2d.5.bn2.weight", "module.layer3_2d.5.bn2.bias", "module.layer3_2d.5.bn2.running_mean", "module.layer3_2d.5.bn2.running_var", "module.layer3_2d.5.bn2.num_batches_tracked", "module.layer4_2d.0.conv1.weight", "module.layer4_2d.0.bn1.weight", "module.layer4_2d.0.bn1.bias", "module.layer4_2d.0.bn1.running_mean", "module.layer4_2d.0.bn1.running_var", "module.layer4_2d.0.bn1.num_batches_tracked", "module.layer4_2d.0.conv2.weight", "module.layer4_2d.0.bn2.weight", "module.layer4_2d.0.bn2.bias", "module.layer4_2d.0.bn2.running_mean", "module.layer4_2d.0.bn2.running_var", "module.layer4_2d.0.bn2.num_batches_tracked", "module.layer4_2d.0.downsample.0.weight", "module.layer4_2d.0.downsample.1.weight", "module.layer4_2d.0.downsample.1.bias", "module.layer4_2d.0.downsample.1.running_mean", "module.layer4_2d.0.downsample.1.running_var", "module.layer4_2d.0.downsample.1.num_batches_tracked", "module.layer4_2d.1.conv1.weight", "module.layer4_2d.1.bn1.weight", "module.layer4_2d.1.bn1.bias", "module.layer4_2d.1.bn1.running_mean", "module.layer4_2d.1.bn1.running_var", "module.layer4_2d.1.bn1.num_batches_tracked", "module.layer4_2d.1.conv2.weight", "module.layer4_2d.1.bn2.weight", "module.layer4_2d.1.bn2.bias", "module.layer4_2d.1.bn2.running_mean", "module.layer4_2d.1.bn2.running_var", "module.layer4_2d.1.bn2.num_batches_tracked", "module.layer4_2d.2.conv1.weight", "module.layer4_2d.2.bn1.weight", "module.layer4_2d.2.bn1.bias", "module.layer4_2d.2.bn1.running_mean", "module.layer4_2d.2.bn1.running_var", "module.layer4_2d.2.bn1.num_batches_tracked", "module.layer4_2d.2.conv2.weight", "module.layer4_2d.2.bn2.weight", "module.layer4_2d.2.bn2.bias", "module.layer4_2d.2.bn2.running_mean", "module.layer4_2d.2.bn2.running_var", "module.layer4_2d.2.bn2.num_batches_tracked", "module.up4_2d.0.weight", "module.up4_2d.0.bias", "module.up4_2d.1.weight", "module.up4_2d.1.bias", "module.up4_2d.1.running_mean", "module.up4_2d.1.running_var", "module.up4_2d.1.num_batches_tracked", "module.delayer4_2d.0.conv1.weight", "module.delayer4_2d.0.bn1.weight", "module.delayer4_2d.0.bn1.bias", "module.delayer4_2d.0.bn1.running_mean", "module.delayer4_2d.0.bn1.running_var", "module.delayer4_2d.0.bn1.num_batches_tracked", "module.delayer4_2d.0.conv2.weight", "module.delayer4_2d.0.bn2.weight", "module.delayer4_2d.0.bn2.bias", "module.delayer4_2d.0.bn2.running_mean", "module.delayer4_2d.0.bn2.running_var", "module.delayer4_2d.0.bn2.num_batches_tracked", "module.delayer4_2d.0.downsample.0.weight", "module.delayer4_2d.0.downsample.1.weight", "module.delayer4_2d.0.downsample.1.bias", "module.delayer4_2d.0.downsample.1.running_mean", "module.delayer4_2d.0.downsample.1.running_var", "module.delayer4_2d.0.downsample.1.num_batches_tracked", "module.delayer4_2d.1.conv1.weight", "module.delayer4_2d.1.bn1.weight", "module.delayer4_2d.1.bn1.bias", "module.delayer4_2d.1.bn1.running_mean", "module.delayer4_2d.1.bn1.running_var", "module.delayer4_2d.1.bn1.num_batches_tracked", "module.delayer4_2d.1.conv2.weight", "module.delayer4_2d.1.bn2.weight", "module.delayer4_2d.1.bn2.bias", "module.delayer4_2d.1.bn2.running_mean", "module.delayer4_2d.1.bn2.running_var", "module.delayer4_2d.1.bn2.num_batches_tracked", "module.delayer4_2d.2.conv1.weight", "module.delayer4_2d.2.bn1.weight", "module.delayer4_2d.2.bn1.bias", "module.delayer4_2d.2.bn1.running_mean", "module.delayer4_2d.2.bn1.running_var", "module.delayer4_2d.2.bn1.num_batches_tracked", "module.delayer4_2d.2.conv2.weight", "module.delayer4_2d.2.bn2.weight", "module.delayer4_2d.2.bn2.bias", "module.delayer4_2d.2.bn2.running_mean", "module.delayer4_2d.2.bn2.running_var", "module.delayer4_2d.2.bn2.num_batches_tracked", "module.up3_2d.0.weight", "module.up3_2d.0.bias", "module.up3_2d.1.weight", "module.up3_2d.1.bias", "module.up3_2d.1.running_mean", "module.up3_2d.1.running_var", "module.up3_2d.1.num_batches_tracked", "module.delayer3_2d.0.conv1.weight", "module.delayer3_2d.0.bn1.weight", "module.delayer3_2d.0.bn1.bias", "module.delayer3_2d.0.bn1.running_mean", "module.delayer3_2d.0.bn1.running_var", "module.delayer3_2d.0.bn1.num_batches_tracked", "module.delayer3_2d.0.conv2.weight", "module.delayer3_2d.0.bn2.weight", "module.delayer3_2d.0.bn2.bias", "module.delayer3_2d.0.bn2.running_mean", "module.delayer3_2d.0.bn2.running_var", "module.delayer3_2d.0.bn2.num_batches_tracked", "module.delayer3_2d.0.downsample.0.weight", "module.delayer3_2d.0.downsample.1.weight", "module.delayer3_2d.0.downsample.1.bias", "module.delayer3_2d.0.downsample.1.running_mean", "module.delayer3_2d.0.downsample.1.running_var", "module.delayer3_2d.0.downsample.1.num_batches_tracked", "module.delayer3_2d.1.conv1.weight", "module.delayer3_2d.1.bn1.weight", "module.delayer3_2d.1.bn1.bias", "module.delayer3_2d.1.bn1.running_mean", "module.delayer3_2d.1.bn1.running_var", "module.delayer3_2d.1.bn1.num_batches_tracked", "module.delayer3_2d.1.conv2.weight", "module.delayer3_2d.1.bn2.weight", "module.delayer3_2d.1.bn2.bias", "module.delayer3_2d.1.bn2.running_mean", "module.delayer3_2d.1.bn2.running_var", "module.delayer3_2d.1.bn2.num_batches_tracked", "module.delayer3_2d.2.conv1.weight", "module.delayer3_2d.2.bn1.weight", "module.delayer3_2d.2.bn1.bias", "module.delayer3_2d.2.bn1.running_mean", "module.delayer3_2d.2.bn1.running_var", "module.delayer3_2d.2.bn1.num_batches_tracked", "module.delayer3_2d.2.conv2.weight", "module.delayer3_2d.2.bn2.weight", "module.delayer3_2d.2.bn2.bias", "module.delayer3_2d.2.bn2.running_mean", "module.delayer3_2d.2.bn2.running_var", "module.delayer3_2d.2.bn2.num_batches_tracked", "module.delayer3_2d.3.conv1.weight", "module.delayer3_2d.3.bn1.weight", "module.delayer3_2d.3.bn1.bias", "module.delayer3_2d.3.bn1.running_mean", "module.delayer3_2d.3.bn1.running_var", "module.delayer3_2d.3.bn1.num_batches_tracked", "module.delayer3_2d.3.conv2.weight", "module.delayer3_2d.3.bn2.weight", "module.delayer3_2d.3.bn2.bias", "module.delayer3_2d.3.bn2.running_mean", "module.delayer3_2d.3.bn2.running_var", "module.delayer3_2d.3.bn2.num_batches_tracked", "module.delayer3_2d.4.conv1.weight", "module.delayer3_2d.4.bn1.weight", "module.delayer3_2d.4.bn1.bias", "module.delayer3_2d.4.bn1.running_mean", "module.delayer3_2d.4.bn1.running_var", "module.delayer3_2d.4.bn1.num_batches_tracked", "module.delayer3_2d.4.conv2.weight", "module.delayer3_2d.4.bn2.weight", "module.delayer3_2d.4.bn2.bias", "module.delayer3_2d.4.bn2.running_mean", "module.delayer3_2d.4.bn2.running_var", "module.delayer3_2d.4.bn2.num_batches_tracked", "module.delayer3_2d.5.conv1.weight", "module.delayer3_2d.5.bn1.weight", "module.delayer3_2d.5.bn1.bias", "module.delayer3_2d.5.bn1.running_mean", "module.delayer3_2d.5.bn1.running_var", "module.delayer3_2d.5.bn1.num_batches_tracked", "module.delayer3_2d.5.conv2.weight", "module.delayer3_2d.5.bn2.weight", "module.delayer3_2d.5.bn2.bias", "module.delayer3_2d.5.bn2.running_mean", "module.delayer3_2d.5.bn2.running_var", "module.delayer3_2d.5.bn2.num_batches_tracked", "module.up2_2d.0.weight", "module.up2_2d.0.bias", "module.up2_2d.1.weight", "module.up2_2d.1.bias", "module.up2_2d.1.running_mean", "module.up2_2d.1.running_var", "module.up2_2d.1.num_batches_tracked", "module.delayer2_2d.0.conv1.weight", "module.delayer2_2d.0.bn1.weight", "module.delayer2_2d.0.bn1.bias", "module.delayer2_2d.0.bn1.running_mean", "module.delayer2_2d.0.bn1.running_var", "module.delayer2_2d.0.bn1.num_batches_tracked", "module.delayer2_2d.0.conv2.weight", "module.delayer2_2d.0.bn2.weight", "module.delayer2_2d.0.bn2.bias", "module.delayer2_2d.0.bn2.running_mean", "module.delayer2_2d.0.bn2.running_var", "module.delayer2_2d.0.bn2.num_batches_tracked", "module.delayer2_2d.0.downsample.0.weight", "module.delayer2_2d.0.downsample.1.weight", "module.delayer2_2d.0.downsample.1.bias", "module.delayer2_2d.0.downsample.1.running_mean", "module.delayer2_2d.0.downsample.1.running_var", "module.delayer2_2d.0.downsample.1.num_batches_tracked", "module.delayer2_2d.1.conv1.weight", "module.delayer2_2d.1.bn1.weight", "module.delayer2_2d.1.bn1.bias", "module.delayer2_2d.1.bn1.running_mean", "module.delayer2_2d.1.bn1.running_var", "module.delayer2_2d.1.bn1.num_batches_tracked", "module.delayer2_2d.1.conv2.weight", "module.delayer2_2d.1.bn2.weight", "module.delayer2_2d.1.bn2.bias", "module.delayer2_2d.1.bn2.running_mean", "module.delayer2_2d.1.bn2.running_var", "module.delayer2_2d.1.bn2.num_batches_tracked", "module.delayer2_2d.2.conv1.weight", "module.delayer2_2d.2.bn1.weight", "module.delayer2_2d.2.bn1.bias", "module.delayer2_2d.2.bn1.running_mean", "module.delayer2_2d.2.bn1.running_var", "module.delayer2_2d.2.bn1.num_batches_tracked", "module.delayer2_2d.2.conv2.weight", "module.delayer2_2d.2.bn2.weight", "module.delayer2_2d.2.bn2.bias", "module.delayer2_2d.2.bn2.running_mean", "module.delayer2_2d.2.bn2.running_var", "module.delayer2_2d.2.bn2.num_batches_tracked", "module.delayer2_2d.3.conv1.weight", "module.delayer2_2d.3.bn1.weight", "module.delayer2_2d.3.bn1.bias", "module.delayer2_2d.3.bn1.running_mean", "module.delayer2_2d.3.bn1.running_var", "module.delayer2_2d.3.bn1.num_batches_tracked", "module.delayer2_2d.3.conv2.weight", "module.delayer2_2d.3.bn2.weight", "module.delayer2_2d.3.bn2.bias", "module.delayer2_2d.3.bn2.running_mean", "module.delayer2_2d.3.bn2.running_var", "module.delayer2_2d.3.bn2.num_batches_tracked", "module.cls_2d.0.weight", "module.cls_2d.1.weight", "module.cls_2d.1.bias", "module.cls_2d.1.running_mean", "module.cls_2d.1.running_var", "module.cls_2d.1.num_batches_tracked", "module.cls_2d.3.weight", "module.cls_2d.3.bias", "module.layer0_3d.0.kernel", "module.layer0_3d.1.bn.weight", "module.layer0_3d.1.bn.bias", "module.layer0_3d.1.bn.running_mean", "module.layer0_3d.1.bn.running_var", "module.layer0_3d.1.bn.num_batches_tracked", "module.layer1_3d.0.kernel", "module.layer1_3d.1.bn.weight", "module.layer1_3d.1.bn.bias", "module.layer1_3d.1.bn.running_mean", "module.layer1_3d.1.bn.running_var", "module.layer1_3d.1.bn.num_batches_tracked", "module.layer1_3d.3.0.conv1.kernel", "module.layer1_3d.3.0.norm1.bn.weight", "module.layer1_3d.3.0.norm1.bn.bias", "module.layer1_3d.3.0.norm1.bn.running_mean", "module.layer1_3d.3.0.norm1.bn.running_var", "module.layer1_3d.3.0.norm1.bn.num_batches_tracked", "module.layer1_3d.3.0.conv2.kernel", "module.layer1_3d.3.0.norm2.bn.weight", "module.layer1_3d.3.0.norm2.bn.bias", "module.layer1_3d.3.0.norm2.bn.running_mean", "module.layer1_3d.3.0.norm2.bn.running_var", "module.layer1_3d.3.0.norm2.bn.num_batches_tracked", "module.layer1_3d.3.1.conv1.kernel", "module.layer1_3d.3.1.norm1.bn.weight", "module.layer1_3d.3.1.norm1.bn.bias", "module.layer1_3d.3.1.norm1.bn.running_mean", "module.layer1_3d.3.1.norm1.bn.running_var", "module.layer1_3d.3.1.norm1.bn.num_batches_tracked", "module.layer1_3d.3.1.conv2.kernel", "module.layer1_3d.3.1.norm2.bn.weight", "module.layer1_3d.3.1.norm2.bn.bias", "module.layer1_3d.3.1.norm2.bn.running_mean", "module.layer1_3d.3.1.norm2.bn.running_var", "module.layer1_3d.3.1.norm2.bn.num_batches_tracked", "module.layer2_3d.0.kernel", "module.layer2_3d.1.bn.weight", "module.layer2_3d.1.bn.bias", "module.layer2_3d.1.bn.running_mean", "module.layer2_3d.1.bn.running_var", "module.layer2_3d.1.bn.num_batches_tracked", "module.layer2_3d.3.0.conv1.kernel", "module.layer2_3d.3.0.norm1.bn.weight", "module.layer2_3d.3.0.norm1.bn.bias", "module.layer2_3d.3.0.norm1.bn.running_mean", "module.layer2_3d.3.0.norm1.bn.running_var", "module.layer2_3d.3.0.norm1.bn.num_batches_tracked", "module.layer2_3d.3.0.conv2.kernel", "module.layer2_3d.3.0.norm2.bn.weight", "module.layer2_3d.3.0.norm2.bn.bias", "module.layer2_3d.3.0.norm2.bn.running_mean", "module.layer2_3d.3.0.norm2.bn.running_var", "module.layer2_3d.3.0.norm2.bn.num_batches_tracked", "module.layer2_3d.3.0.downsample.0.kernel", "module.layer2_3d.3.0.downsample.1.bn.weight", "module.layer2_3d.3.0.downsample.1.bn.bias", "module.layer2_3d.3.0.downsample.1.bn.running_mean", "module.layer2_3d.3.0.downsample.1.bn.running_var", "module.layer2_3d.3.0.downsample.1.bn.num_batches_tracked", "module.layer2_3d.3.1.conv1.kernel", "module.layer2_3d.3.1.norm1.bn.weight", "module.layer2_3d.3.1.norm1.bn.bias", "module.layer2_3d.3.1.norm1.bn.running_mean", "module.layer2_3d.3.1.norm1.bn.running_var", "module.layer2_3d.3.1.norm1.bn.num_batches_tracked", "module.layer2_3d.3.1.conv2.kernel", "module.layer2_3d.3.1.norm2.bn.weight", "module.layer2_3d.3.1.norm2.bn.bias", "module.layer2_3d.3.1.norm2.bn.running_mean", "module.layer2_3d.3.1.norm2.bn.running_var", "module.layer2_3d.3.1.norm2.bn.num_batches_tracked", "module.layer3_3d.0.kernel", "module.layer3_3d.1.bn.weight", "module.layer3_3d.1.bn.bias", "module.layer3_3d.1.bn.running_mean", "module.layer3_3d.1.bn.running_var", "module.layer3_3d.1.bn.num_batches_tracked", "module.layer3_3d.3.0.conv1.kernel", "module.layer3_3d.3.0.norm1.bn.weight", "module.layer3_3d.3.0.norm1.bn.bias", "module.layer3_3d.3.0.norm1.bn.running_mean", "module.layer3_3d.3.0.norm1.bn.running_var", "module.layer3_3d.3.0.norm1.bn.num_batches_tracked", "module.layer3_3d.3.0.conv2.kernel", "module.layer3_3d.3.0.norm2.bn.weight", "module.layer3_3d.3.0.norm2.bn.bias", "module.layer3_3d.3.0.norm2.bn.running_mean", "module.layer3_3d.3.0.norm2.bn.running_var", "module.layer3_3d.3.0.norm2.bn.num_batches_tracked", "module.layer3_3d.3.0.downsample.0.kernel", "module.layer3_3d.3.0.downsample.1.bn.weight", "module.layer3_3d.3.0.downsample.1.bn.bias", "module.layer3_3d.3.0.downsample.1.bn.running_mean", "module.layer3_3d.3.0.downsample.1.bn.running_var", "module.layer3_3d.3.0.downsample.1.bn.num_batches_tracked", "module.layer3_3d.3.1.conv1.kernel", "module.layer3_3d.3.1.norm1.bn.weight", "module.layer3_3d.3.1.norm1.bn.bias", "module.layer3_3d.3.1.norm1.bn.running_mean", "module.layer3_3d.3.1.norm1.bn.running_var", "module.layer3_3d.3.1.norm1.bn.num_batches_tracked", "module.layer3_3d.3.1.conv2.kernel", "module.layer3_3d.3.1.norm2.bn.weight", "module.layer3_3d.3.1.norm2.bn.bias", "module.layer3_3d.3.1.norm2.bn.running_mean", "module.layer3_3d.3.1.norm2.bn.running_var", "module.layer3_3d.3.1.norm2.bn.num_batches_tracked", "module.layer4_3d.0.kernel", "module.layer4_3d.1.bn.weight", "module.layer4_3d.1.bn.bias", "module.layer4_3d.1.bn.running_mean", "module.layer4_3d.1.bn.running_var", "module.layer4_3d.1.bn.num_batches_tracked", "module.layer4_3d.3.0.conv1.kernel", "module.layer4_3d.3.0.norm1.bn.weight", "module.layer4_3d.3.0.norm1.bn.bias", "module.layer4_3d.3.0.norm1.bn.running_mean", "module.layer4_3d.3.0.norm1.bn.running_var", "module.layer4_3d.3.0.norm1.bn.num_batches_tracked", "module.layer4_3d.3.0.conv2.kernel", "module.layer4_3d.3.0.norm2.bn.weight", "module.layer4_3d.3.0.norm2.bn.bias", "module.layer4_3d.3.0.norm2.bn.running_mean", "module.layer4_3d.3.0.norm2.bn.running_var", "module.layer4_3d.3.0.norm2.bn.num_batches_tracked", "module.layer4_3d.3.0.downsample.0.kernel", "module.layer4_3d.3.0.downsample.1.bn.weight", "module.layer4_3d.3.0.downsample.1.bn.bias", "module.layer4_3d.3.0.downsample.1.bn.running_mean", "module.layer4_3d.3.0.downsample.1.bn.running_var", "module.layer4_3d.3.0.downsample.1.bn.num_batches_tracked", "module.layer4_3d.3.1.conv1.kernel", "module.layer4_3d.3.1.norm1.bn.weight", "module.layer4_3d.3.1.norm1.bn.bias", "module.layer4_3d.3.1.norm1.bn.running_mean", "module.layer4_3d.3.1.norm1.bn.running_var", "module.layer4_3d.3.1.norm1.bn.num_batches_tracked", "module.layer4_3d.3.1.conv2.kernel", "module.layer4_3d.3.1.norm2.bn.weight", "module.layer4_3d.3.1.norm2.bn.bias", "module.layer4_3d.3.1.norm2.bn.running_mean", "module.layer4_3d.3.1.norm2.bn.running_var", "module.layer4_3d.3.1.norm2.bn.num_batches_tracked", "module.layer5_3d.0.kernel", "module.layer5_3d.1.bn.weight", "module.layer5_3d.1.bn.bias", "module.layer5_3d.1.bn.running_mean", "module.layer5_3d.1.bn.running_var", "module.layer5_3d.1.bn.num_batches_tracked", "module.layer6_3d.0.0.conv1.kernel", "module.layer6_3d.0.0.norm1.bn.weight", "module.layer6_3d.0.0.norm1.bn.bias", "module.layer6_3d.0.0.norm1.bn.running_mean", "module.layer6_3d.0.0.norm1.bn.running_var", "module.layer6_3d.0.0.norm1.bn.num_batches_tracked", "module.layer6_3d.0.0.conv2.kernel", "module.layer6_3d.0.0.norm2.bn.weight", "module.layer6_3d.0.0.norm2.bn.bias", "module.layer6_3d.0.0.norm2.bn.running_mean", "module.layer6_3d.0.0.norm2.bn.running_var", "module.layer6_3d.0.0.norm2.bn.num_batches_tracked", "module.layer6_3d.0.0.downsample.0.kernel", "module.layer6_3d.0.0.downsample.1.bn.weight", "module.layer6_3d.0.0.downsample.1.bn.bias", "module.layer6_3d.0.0.downsample.1.bn.running_mean", "module.layer6_3d.0.0.downsample.1.bn.running_var", "module.layer6_3d.0.0.downsample.1.bn.num_batches_tracked", "module.layer6_3d.0.1.conv1.kernel", "module.layer6_3d.0.1.norm1.bn.weight", "module.layer6_3d.0.1.norm1.bn.bias", "module.layer6_3d.0.1.norm1.bn.running_mean", "module.layer6_3d.0.1.norm1.bn.running_var", "module.layer6_3d.0.1.norm1.bn.num_batches_tracked", "module.layer6_3d.0.1.conv2.kernel", "module.layer6_3d.0.1.norm2.bn.weight", "module.layer6_3d.0.1.norm2.bn.bias", "module.layer6_3d.0.1.norm2.bn.running_mean", "module.layer6_3d.0.1.norm2.bn.running_var", "module.layer6_3d.0.1.norm2.bn.num_batches_tracked", "module.layer6_3d.1.kernel", "module.layer6_3d.2.bn.weight", "module.layer6_3d.2.bn.bias", "module.layer6_3d.2.bn.running_mean", "module.layer6_3d.2.bn.running_var", "module.layer6_3d.2.bn.num_batches_tracked", "module.layer7_3d.0.0.conv1.kernel", "module.layer7_3d.0.0.norm1.bn.weight", "module.layer7_3d.0.0.norm1.bn.bias", "module.layer7_3d.0.0.norm1.bn.running_mean", "module.layer7_3d.0.0.norm1.bn.running_var", "module.layer7_3d.0.0.norm1.bn.num_batches_tracked", "module.layer7_3d.0.0.conv2.kernel", "module.layer7_3d.0.0.norm2.bn.weight", "module.layer7_3d.0.0.norm2.bn.bias", "module.layer7_3d.0.0.norm2.bn.running_mean", "module.layer7_3d.0.0.norm2.bn.running_var", "module.layer7_3d.0.0.norm2.bn.num_batches_tracked", "module.layer7_3d.0.0.downsample.0.kernel", "module.layer7_3d.0.0.downsample.1.bn.weight", "module.layer7_3d.0.0.downsample.1.bn.bias", "module.layer7_3d.0.0.downsample.1.bn.running_mean", "module.layer7_3d.0.0.downsample.1.bn.running_var", "module.layer7_3d.0.0.downsample.1.bn.num_batches_tracked", "module.layer7_3d.0.1.conv1.kernel", "module.layer7_3d.0.1.norm1.bn.weight", "module.layer7_3d.0.1.norm1.bn.bias", "module.layer7_3d.0.1.norm1.bn.running_mean", "module.layer7_3d.0.1.norm1.bn.running_var", "module.layer7_3d.0.1.norm1.bn.num_batches_tracked", "module.layer7_3d.0.1.conv2.kernel", "module.layer7_3d.0.1.norm2.bn.weight", "module.layer7_3d.0.1.norm2.bn.bias", "module.layer7_3d.0.1.norm2.bn.running_mean", "module.layer7_3d.0.1.norm2.bn.running_var", "module.layer7_3d.0.1.norm2.bn.num_batches_tracked", "module.layer7_3d.1.kernel", "module.layer7_3d.2.bn.weight", "module.layer7_3d.2.bn.bias", "module.layer7_3d.2.bn.running_mean", "module.layer7_3d.2.bn.running_var", "module.layer7_3d.2.bn.num_batches_tracked", "module.layer8_3d.0.0.conv1.kernel", "module.layer8_3d.0.0.norm1.bn.weight", "module.layer8_3d.0.0.norm1.bn.bias", "module.layer8_3d.0.0.norm1.bn.running_mean", "module.layer8_3d.0.0.norm1.bn.running_var", "module.layer8_3d.0.0.norm1.bn.num_batches_tracked", "module.layer8_3d.0.0.conv2.kernel", "module.layer8_3d.0.0.norm2.bn.weight", "module.layer8_3d.0.0.norm2.bn.bias", "module.layer8_3d.0.0.norm2.bn.running_mean", "module.layer8_3d.0.0.norm2.bn.running_var", "module.layer8_3d.0.0.norm2.bn.num_batches_tracked", "module.layer8_3d.0.0.downsample.0.kernel", "module.layer8_3d.0.0.downsample.1.bn.weight", "module.layer8_3d.0.0.downsample.1.bn.bias", "module.layer8_3d.0.0.downsample.1.bn.running_mean", "module.layer8_3d.0.0.downsample.1.bn.running_var", "module.layer8_3d.0.0.downsample.1.bn.num_batches_tracked", "module.layer8_3d.0.1.conv1.kernel", "module.layer8_3d.0.1.norm1.bn.weight", "module.layer8_3d.0.1.norm1.bn.bias", "module.layer8_3d.0.1.norm1.bn.running_mean", "module.layer8_3d.0.1.norm1.bn.running_var", "module.layer8_3d.0.1.norm1.bn.num_batches_tracked", "module.layer8_3d.0.1.conv2.kernel", "module.layer8_3d.0.1.norm2.bn.weight", "module.layer8_3d.0.1.norm2.bn.bias", "module.layer8_3d.0.1.norm2.bn.running_mean", "module.layer8_3d.0.1.norm2.bn.running_var", "module.layer8_3d.0.1.norm2.bn.num_batches_tracked", "module.layer8_3d.1.kernel", "module.layer8_3d.2.bn.weight", "module.layer8_3d.2.bn.bias", "module.layer8_3d.2.bn.running_mean", "module.layer8_3d.2.bn.running_var", "module.layer8_3d.2.bn.num_batches_tracked", "module.layer9_3d.0.conv1.kernel", "module.layer9_3d.0.norm1.bn.weight", "module.layer9_3d.0.norm1.bn.bias", "module.layer9_3d.0.norm1.bn.running_mean", "module.layer9_3d.0.norm1.bn.running_var", "module.layer9_3d.0.norm1.bn.num_batches_tracked", "module.layer9_3d.0.conv2.kernel", "module.layer9_3d.0.norm2.bn.weight", "module.layer9_3d.0.norm2.bn.bias", "module.layer9_3d.0.norm2.bn.running_mean", "module.layer9_3d.0.norm2.bn.running_var", "module.layer9_3d.0.norm2.bn.num_batches_tracked", "module.layer9_3d.0.downsample.0.kernel", "module.layer9_3d.0.downsample.1.bn.weight", "module.layer9_3d.0.downsample.1.bn.bias", "module.layer9_3d.0.downsample.1.bn.running_mean", "module.layer9_3d.0.downsample.1.bn.running_var", "module.layer9_3d.0.downsample.1.bn.num_batches_tracked", "module.layer9_3d.1.conv1.kernel", "module.layer9_3d.1.norm1.bn.weight", "module.layer9_3d.1.norm1.bn.bias", "module.layer9_3d.1.norm1.bn.running_mean", "module.layer9_3d.1.norm1.bn.running_var", "module.layer9_3d.1.norm1.bn.num_batches_tracked", "module.layer9_3d.1.conv2.kernel", "module.layer9_3d.1.norm2.bn.weight", "module.layer9_3d.1.norm2.bn.bias", "module.layer9_3d.1.norm2.bn.running_mean", "module.layer9_3d.1.norm2.bn.running_var", "module.layer9_3d.1.norm2.bn.num_batches_tracked", "module.cls_3d.kernel", "module.cls_3d.bias", "module.linker_p2.view_fusion.0.kernel", "module.linker_p2.view_fusion.1.bn.weight", "module.linker_p2.view_fusion.1.bn.bias", "module.linker_p2.view_fusion.1.bn.running_mean", "module.linker_p2.view_fusion.1.bn.running_var", "module.linker_p2.view_fusion.1.bn.num_batches_tracked", "module.linker_p2.view_fusion.3.kernel", "module.linker_p2.view_fusion.4.bn.weight", "module.linker_p2.view_fusion.4.bn.bias", "module.linker_p2.view_fusion.4.bn.running_mean", "module.linker_p2.view_fusion.4.bn.running_var", "module.linker_p2.view_fusion.4.bn.num_batches_tracked", "module.linker_p2.fuseTo3d.0.kernel", "module.linker_p2.fuseTo3d.1.bn.weight", "module.linker_p2.fuseTo3d.1.bn.bias", "module.linker_p2.fuseTo3d.1.bn.running_mean", "module.linker_p2.fuseTo3d.1.bn.running_var", "module.linker_p2.fuseTo3d.1.bn.num_batches_tracked", "module.linker_p2.view_sep.0.kernel", "module.linker_p2.view_sep.1.bn.weight", "module.linker_p2.view_sep.1.bn.bias", "module.linker_p2.view_sep.1.bn.running_mean", "module.linker_p2.view_sep.1.bn.running_var", "module.linker_p2.view_sep.1.bn.num_batches_tracked", "module.linker_p2.fuseTo2d.0.weight", "module.linker_p2.fuseTo2d.1.weight", "module.linker_p2.fuseTo2d.1.bias", "module.linker_p2.fuseTo2d.1.running_mean", "module.linker_p2.fuseTo2d.1.running_var", "module.linker_p2.fuseTo2d.1.num_batches_tracked", "module.linker_p3.view_fusion.0.kernel", "module.linker_p3.view_fusion.1.bn.weight", "module.linker_p3.view_fusion.1.bn.bias", "module.linker_p3.view_fusion.1.bn.running_mean", "module.linker_p3.view_fusion.1.bn.running_var", "module.linker_p3.view_fusion.1.bn.num_batches_tracked", "module.linker_p3.view_fusion.3.kernel", "module.linker_p3.view_fusion.4.bn.weight", "module.linker_p3.view_fusion.4.bn.bias", "module.linker_p3.view_fusion.4.bn.running_mean", "module.linker_p3.view_fusion.4.bn.running_var", "module.linker_p3.view_fusion.4.bn.num_batches_tracked", "module.linker_p3.fuseTo3d.0.kernel", "module.linker_p3.fuseTo3d.1.bn.weight", "module.linker_p3.fuseTo3d.1.bn.bias", "module.linker_p3.fuseTo3d.1.bn.running_mean", "module.linker_p3.fuseTo3d.1.bn.running_var", "module.linker_p3.fuseTo3d.1.bn.num_batches_tracked", "module.linker_p3.view_sep.0.kernel", "module.linker_p3.view_sep.1.bn.weight", "module.linker_p3.view_sep.1.bn.bias", "module.linker_p3.view_sep.1.bn.running_mean", "module.linker_p3.view_sep.1.bn.running_var", "module.linker_p3.view_sep.1.bn.num_batches_tracked", "module.linker_p3.fuseTo2d.0.weight", "module.linker_p3.fuseTo2d.1.weight", "module.linker_p3.fuseTo2d.1.bias", "module.linker_p3.fuseTo2d.1.running_mean", "module.linker_p3.fuseTo2d.1.running_var", "module.linker_p3.fuseTo2d.1.num_batches_tracked", "module.linker_p4.view_fusion.0.kernel", "module.linker_p4.view_fusion.1.bn.weight", "module.linker_p4.view_fusion.1.bn.bias", "module.linker_p4.view_fusion.1.bn.running_mean", "module.linker_p4.view_fusion.1.bn.running_var", "module.linker_p4.view_fusion.1.bn.num_batches_tracked", "module.linker_p4.view_fusion.3.kernel", "module.linker_p4.view_fusion.4.bn.weight", "module.linker_p4.view_fusion.4.bn.bias", "module.linker_p4.view_fusion.4.bn.running_mean", "module.linker_p4.view_fusion.4.bn.running_var", "module.linker_p4.view_fusion.4.bn.num_batches_tracked", "module.linker_p4.fuseTo3d.0.kernel", "module.linker_p4.fuseTo3d.1.bn.weight", "module.linker_p4.fuseTo3d.1.bn.bias", "module.linker_p4.fuseTo3d.1.bn.running_mean", "module.linker_p4.fuseTo3d.1.bn.running_var", "module.linker_p4.fuseTo3d.1.bn.num_batches_tracked", "module.linker_p4.view_sep.0.kernel", "module.linker_p4.view_sep.1.bn.weight", "module.linker_p4.view_sep.1.bn.bias", "module.linker_p4.view_sep.1.bn.running_mean", "module.linker_p4.view_sep.1.bn.running_var", "module.linker_p4.view_sep.1.bn.num_batches_tracked", "module.linker_p4.fuseTo2d.0.weight", "module.linker_p4.fuseTo2d.1.weight", "module.linker_p4.fuseTo2d.1.bias", "module.linker_p4.fuseTo2d.1.running_mean", "module.linker_p4.fuseTo2d.1.running_var", "module.linker_p4.fuseTo2d.1.num_batches_tracked", "module.linker_p5.view_fusion.0.kernel", "module.linker_p5.view_fusion.1.bn.weight", "module.linker_p5.view_fusion.1.bn.bias", "module.linker_p5.view_fusion.1.bn.running_mean", "module.linker_p5.view_fusion.1.bn.running_var", "module.linker_p5.view_fusion.1.bn.num_batches_tracked", "module.linker_p5.view_fusion.3.kernel", "module.linker_p5.view_fusion.4.bn.weight", "module.linker_p5.view_fusion.4.bn.bias", "module.linker_p5.view_fusion.4.bn.running_mean", "module.linker_p5.view_fusion.4.bn.running_var", "module.linker_p5.view_fusion.4.bn.num_batches_tracked", "module.linker_p5.fuseTo3d.0.kernel", "module.linker_p5.fuseTo3d.1.bn.weight", "module.linker_p5.fuseTo3d.1.bn.bias", "module.linker_p5.fuseTo3d.1.bn.running_mean", "module.linker_p5.fuseTo3d.1.bn.running_var", "module.linker_p5.fuseTo3d.1.bn.num_batches_tracked", "module.linker_p5.view_sep.0.kernel", "module.linker_p5.view_sep.1.bn.weight", "module.linker_p5.view_sep.1.bn.bias", "module.linker_p5.view_sep.1.bn.running_mean", "module.linker_p5.view_sep.1.bn.running_var", "module.linker_p5.view_sep.1.bn.num_batches_tracked", "module.linker_p5.fuseTo2d.0.weight", "module.linker_p5.fuseTo2d.1.weight", "module.linker_p5.fuseTo2d.1.bias", "module.linker_p5.fuseTo2d.1.running_mean", "module.linker_p5.fuseTo2d.1.running_var", "module.linker_p5.fuseTo2d.1.num_batches_tracked".

I've already solved too many problems to get this error. I sure even if i solve this error, another error will occur. My conclusion is that you have not tested this on other environment at all. Please test your opensource on other environment. And give more details on README.md about how to run this.

Thank you.

First of all, thank you for your wonderful work! Now, I normally try to reproduce the 3D semantic segmentation results of MinkowskiUNet18A in the paper, but it seems that the accuracy can only reach 55%, which is 13% different from the 68% shown in Table 5 in the paper. Is there any problem with the training process?

The following are my experimental parameter settings：

Thank you again for your excellent work and look forward to your reply. Best regards!

Hi! Thanks for your great work and source code!

I tried to run the code but met several problems. I am confused about the data structure of scannet in this repo. As there is some pieces of code like

for x in self.data_paths: 
    ps = glob(join(x[:-15].replace(split, '2D'), 'color', '*.jpg'))

Could you give a specific explanation for easily usage? Thanks a lot!

Your paper is excellent. The idea of BPM is beautiful. and Explanation about data structure of BPM is very kind and easy-to-understand. Thank you for your contribution to this field.

But after reading your paper, I still have some points which i can't understand, and I think if some explanations for these are provided, it will be great improvements for me, and for your paper as well!

First, I cannot understand the philosophy behind the method to get point cloud as input of the 3D network. Why extract 3D point cloud from reconstructed 3D surface? Why don't you create point cloud from depth frame directly? Your said, "as done in [Mincowski Paper]". But, Even after reading [Mincowski Paper], I cannot have any idea about why extract 3D point cloud from reconstructed 3D surface. Can you explain it?

Second, I want to point out the ablation study with regard to the ablation for the number of 2D views. The test on 2D mIoU is OK. But, To the best of my understanding about your paper, The test result on 3D mIoU is very confusing. Let me give a example situation. If you have 1 scene and there are 5 views, And about 5 number of view result, It's OK. And then, About 4 views result, What's going on? Isn't there some Not-Evaluated part on that scene? Or, To evaluate whole scene by 4 views method, There will be overlapped voxels inevitably. How to evaluate these overlaps fairly?

Third, There will be both overlapped area and non-overlapped area on ScanNet Dataset even after view sampling. I think evaluating only on mixed(overlapped area and non-overlapped) parts is not a fully ablation. I suggest you to give ablation result(2D/3D mIoU) only with 'overlapped parts', separately with mixed result.

Thank you.

Hi, What an excellent work!

You mentioned "We train our BPNet on the ScanNetV2 dataset for 100 epochs" in the paper. And I am interested in the details about the 100 epochs training. For example, what backbone (e.g., ResNet34 from the config file), how many GPUs, what kind of GPUs (e.g., V100?), and the total training hours~

I would appreciate it very much if you can share these details~ Thanks~

RT, we are focusing on the efficiency of 3d models, so we want to know how was the time performance for bpnet. BTW, thanks for your talent work, it is fabulous

Hey, @wbhu !

I would like to ask why the train() function in train.py doesn't have dist.all_reduce funciton but validate() has?

the function like this in train.py as follows:

if args.multiprocessing_distributed:
    dist.all_reduce(intersection), dist.all_reduce(union), dist.all_reduce(target)

I submitted it according to the format for 3d semantic segmentation. I don't know where the processing occurs. Please tell me what these four files are respectively？ Thanks！！！

Dear benchmark user,

Unfortunately, our evaluation system failed to evaluate your result submission for ****. Reason: results folder does not contain 4 subfolders with correct format:

The system thinks that the problem is due to invalid input. However, if you think it is the system's fault, please contact an administrator.

Best regards, The ScanNet Benchmark administrators

I change the data_name 'scannet_cross' to 'scannet_3d_mink' in yaml , then I got this error when I ran train.py. I find the model for 'scannet_3d_mink' was not be written.

https://github.com/wbhu/BPNet/blob/efba4b15998e03fbfb15a60ce333bfbde489d8ba/config/scannet/bpnet_5cm.yaml#L24

What is the purpose of using this loop argument compared to just increasing the number of epochs?

Hi,

the ScanNetV2 dataset is too large, so I only want to download the needed parts of this dataset. Are the file types that BPNet uses .sens/_vh_clean_2.ply/_vh_clean_2.labels.ply/_2d-label.zip, and e.g. run 'python download-scannet.py -o [directory in which to download] --type .sens'? If not, could you tell me which parts of ScanNet V2 need to be downloaded? Thanks!

the entire valid file suffixes are:

<scanId>
|-- <scanId>.sens
    RGB-D sensor stream containing color frames, depth frames, camera poses and other data
|-- <scanId>_vh_clean.ply
    High quality reconstructed mesh
|-- <scanId>_vh_clean_2.ply
    Cleaned and decimated mesh for semantic annotations
|-- <scanId>_vh_clean_2.0.010000.segs.json
    Over-segmentation of annotation mesh
|-- <scanId>.aggregation.json, <scanId>_vh_clean.aggregation.json
    Aggregated instance-level semantic annotations on lo-res, hi-res meshes, respectively
|-- <scanId>_vh_clean_2.0.010000.segs.json, <scanId>_vh_clean.segs.json
    Over-segmentation of lo-res, hi-res meshes, respectively (referenced by aggregated semantic annotations)
|-- <scanId>_vh_clean_2.labels.ply
    Visualization of aggregated semantic segmentation; colored by nyu40 labels (see img/legend; ply property 'label' denotes the ScanNet label id)
|-- <scanId>_2d-label.zip
    Raw 2d projections of aggregated annotation labels as 16-bit pngs with ScanNet label ids
|-- <scanId>_2d-instance.zip
    Raw 2d projections of aggregated annotation instances as 8-bit pngs
|-- <scanId>_2d-label-filt.zip
    Filtered 2d projections of aggregated annotation labels as 16-bit pngs with ScanNet label ids
|-- <scanId>_2d-instance-filt.zip
    Filtered 2d projections of aggregated annotation instances as 8-bit pngs

hi，i follow the instruction and after i git checkout XXX and use this bash "python setup.py install"

there exists some error cannot find -lminkowski: No such file or directory

any help would be apperatiated,thanks.

and on ubuntu20 my nvcc version is 11.4. and i strictly follow the instructions.

I reproduced your source code with no other substantial changes, and with the parameter settings of UNet34 for the 2D network and Mink18A for the 3D network and voxel size=5cm, the 3D mIOU basically achieves the performance of 70.6 reported in your paper. However, the 2D mIOU is only 47.3, while your paper's result is 65.1, is this something I'm missing？

Hello, i try to run prepare_2d_data but found

if opt.export_label_images:
    assert opt.label_map_file != ''

and it need path to scannetv2-labels.combined.tsv ,where is it? Thanks for the code

Thank you very much for your excellent work, I had the following problems in reading your training code and test code： In 'train.py' file, val_sampler=torch.utils.data.distributed.DistributedSampler(val_data) , with dist.all_reduce(intersection), dist.all_reduce(union), dist.all_reduce(target). In 'test.py' file, val_sampler=None, with dist.all_reduce(output_3d).

My question：

1. Why is the sampler inconsistent here?
2. I found that performance did not change when val_sampler=None. What is the significance of dist.all_reduce() here?
3. I found that when val_sampler=torch.utils.data.distributed.DistributedSampler(val_data) and dist.all_reduce() was not used, mIOU increased during testing. Why is this?
4. The last question is that in the 'train.py' file, the "intersectionAndUnionGPU" function in the ‘util.py’ file is used, while in the 'test.py' file, the ‘evaluate’ function in the ‘iou.py’ file is used. What are the essential differences between the two evaluation metrics in terms of application?

I look forward to hearing from you and thank you again for your excellent work.

Hello, thanks for sharing your code.

I wanted to reproduce the 2D miou score given in your article (61.5 miou for the 2D only) but i did not manage to go over 56.2 miou.

I used your preprocessing script for 2d datas, a resnet34 similar to yours (using conv2dTranspose over F.interpolate) and when i tried using the same data augmentation (random crop on image with gaussian blur on training set), the network miou score worsen.

Did i miss something ? Thanks for your answer