2026/03/14 22:31:06 - bevformer - INFO - 
------------------------------------------------------------
System environment:
    sys.platform: darwin
    Python: 3.10.0 (default, Oct 17 2021, 11:56:26) [Clang 13.0.0 ]
    CUDA available: False
    numpy_random_seed: 321
    GCC: Apple clang version 15.0.0 (clang-1500.3.9.4)
    PyTorch: 2.10.0
    PyTorch compiling details: PyTorch built with:
  - GCC 4.2
  - C++ Version: 201703
  - clang 15.0.0
  - OpenMP 201811
  - LAPACK is enabled (usually provided by MKL)
  - NNPACK is enabled
  - CPU capability usage: DEFAULT
  - Build settings: BLAS_INFO=accelerate, BUILD_TYPE=Release, COMMIT_SHA=449b1768410104d3ed79d3bcfe4ba1d65c7f22c0, CXX_COMPILER=/usr/bin/c++, CXX_FLAGS= -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DLIBKINETO_NOXPUPTI=ON -DUSE_PYTORCH_QNNPACK -DAT_BUILD_ARM_VEC256_WITH_SLEEF -DUSE_XNNPACK -DUSE_PYTORCH_METAL_EXPORT -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DUSE_COREML_DELEGATE -O2 -fPIC -DC10_NODEPRECATED -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=braced-scalar-init -Werror=range-loop-construct -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-unknown-pragmas -Wno-unused-parameter -Wno-strict-overflow -Wno-strict-aliasing -Wvla-extension -Wsuggest-override -Wnewline-eof -Winconsistent-missing-override -Winconsistent-missing-destructor-override -Wno-pass-failed -Wno-error=old-style-cast -Wconstant-conversion -Qunused-arguments -faligned-new -fno-math-errno -fno-trapping-math -Werror=format -DUSE_MPS -Wno-missing-braces, LAPACK_INFO=accelerate, TORCH_VERSION=2.10.0, USE_CUDA=OFF, USE_CUDNN=OFF, USE_CUSPARSELT=OFF, USE_EIGEN_FOR_BLAS=ON, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=ON, USE_MKL=OFF, USE_MKLDNN=OFF, USE_MPI=OFF, USE_NCCL=OFF, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF, USE_XCCL=OFF, USE_XPU=OFF, 

    TorchVision: 0.25.0
    OpenCV: 4.13.0
    MMEngine: 0.8.5

Runtime environment:
    dist_cfg: {'backend': 'nccl'}
    seed: 321
    Distributed launcher: none
    Distributed training: False
    GPU number: 1
------------------------------------------------------------

2026/03/14 22:31:06 - bevformer - INFO - Set random seed to 321, deterministic: False
2026/03/14 22:31:06 - bevformer - INFO - Building model
2026/03/14 22:31:07 - bevformer - INFO - Model architecture:
BEVFormerDetector(
  (data_preprocessor): BaseDataPreprocessor()
  (pts_bbox_head): BEVFormerHead(
    (loss_cls): FocalLoss()
    (loss_bbox): L1Loss()
    (loss_iou): GIoULoss()
    (transformer): PerceptionTransformer(
      (encoder): BEVFormerEncoder(
        (layers): ModuleList(
          (0-2): 3 x BEVFormerLayer(
            (attentions): ModuleList(
              (0): TemporalSelfAttention(
                (sampling_offsets): Linear(in_features=512, out_features=128, bias=True)
                (attention_weights): Linear(in_features=512, out_features=64, bias=True)
                (value_proj): Linear(in_features=256, out_features=256, bias=True)
                (output_proj): Linear(in_features=256, out_features=256, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (1): SpatialCrossAttention(
                (deformable_attention): MSDeformableAttention3D(
                  (sampling_offsets): Linear(in_features=256, out_features=128, bias=True)
                  (attention_weights): Linear(in_features=256, out_features=64, bias=True)
                  (value_proj): Linear(in_features=256, out_features=256, bias=True)
                )
                (output_proj): Linear(in_features=256, out_features=256, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (ffns): ModuleList(
              (0): FFN(
                (layers): Sequential(
                  (0): Sequential(
                    (0): Linear(in_features=256, out_features=512, bias=True)
                    (1): ReLU(inplace=True)
                    (2): Dropout(p=0.1, inplace=False)
                  )
                  (1): Linear(in_features=512, out_features=256, bias=True)
                  (2): Dropout(p=0.1, inplace=False)
                )
                (dropout_layer): Identity()
                (gamma2): Identity()
              )
            )
            (norms): ModuleList(
              (0-2): 3 x LayerNorm((256,), eps=1e-05, elementwise_affine=True)
            )
          )
        )
      )
      (decoder): DetectionTransformerDecoder(
        (layers): ModuleList(
          (0-5): 6 x DetrTransformerDecoderLayer(
            (attentions): ModuleList(
              (0): MultiheadAttention(
                (q_proj): Linear(in_features=256, out_features=256, bias=True)
                (k_proj): Linear(in_features=256, out_features=256, bias=True)
                (v_proj): Linear(in_features=256, out_features=256, bias=True)
                (output_proj): Linear(in_features=256, out_features=256, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (1): CustomMSDeformableAttention(
                (sampling_offsets): Linear(in_features=256, out_features=64, bias=True)
                (attention_weights): Linear(in_features=256, out_features=32, bias=True)
                (value_proj): Linear(in_features=256, out_features=256, bias=True)
                (output_proj): Linear(in_features=256, out_features=256, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (ffns): ModuleList(
              (0): FFN(
                (layers): Sequential(
                  (0): Sequential(
                    (0): Linear(in_features=256, out_features=512, bias=True)
                    (1): ReLU(inplace=True)
                    (2): Dropout(p=0.1, inplace=False)
                  )
                  (1): Linear(in_features=512, out_features=256, bias=True)
                  (2): Dropout(p=0.1, inplace=False)
                )
                (dropout_layer): Identity()
                (gamma2): Identity()
              )
            )
            (norms): ModuleList(
              (0-2): 3 x LayerNorm((256,), eps=1e-05, elementwise_affine=True)
            )
          )
        )
      )
      (reference_points): Linear(in_features=256, out_features=3, bias=True)
      (can_bus_mlp): Sequential(
        (0): Linear(in_features=18, out_features=128, bias=True)
        (1): ReLU(inplace=True)
        (2): Linear(in_features=128, out_features=256, bias=True)
        (3): ReLU(inplace=True)
        (norm): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
      )
    )
    (cls_branches): ModuleList(
      (0-5): 6 x Sequential(
        (0): Linear(in_features=256, out_features=256, bias=True)
        (1): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
        (2): ReLU(inplace=True)
        (3): Linear(in_features=256, out_features=10, bias=True)
      )
    )
    (reg_branches): ModuleList(
      (0-5): 6 x Sequential(
        (0): Linear(in_features=256, out_features=256, bias=True)
        (1): ReLU()
        (2): Linear(in_features=256, out_features=256, bias=True)
        (3): ReLU()
        (4): Linear(in_features=256, out_features=10, bias=True)
      )
    )
    (bev_embedding): Embedding(2500, 256)
    (object_query_embedding): Embedding(900, 512)
    (positional_encoding): LearnedPositionalEncoding(num_feats=128, row_num_embed=50, col_num_embed=50)
  )
  (img_backbone): ResNet(
    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): ReLU(inplace=True)
    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
    (layer1): ResLayer(
      (0): Bottleneck(
        (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (downsample): Sequential(
          (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (1): Bottleneck(
        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (2): Bottleneck(
        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
    )
    (layer2): ResLayer(
      (0): Bottleneck(
        (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (downsample): Sequential(
          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (1): Bottleneck(
        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (2): Bottleneck(
        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (3): Bottleneck(
        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
    )
    (layer3): ResLayer(
      (0): Bottleneck(
        (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (downsample): Sequential(
          (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
          (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (1): Bottleneck(
        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (2): Bottleneck(
        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (3): Bottleneck(
        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (4): Bottleneck(
        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (5): Bottleneck(
        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
    )
    (layer4): ResLayer(
      (0): Bottleneck(
        (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
        (downsample): Sequential(
          (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
          (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (1): Bottleneck(
        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
      (2): Bottleneck(
        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu): ReLU(inplace=True)
      )
      init_cfg={'type': 'Constant', 'val': 0, 'override': {'name': 'norm3'}}
    )
  )
  init_cfg=[{'type': 'Kaiming', 'layer': 'Conv2d'}, {'type': 'Constant', 'val': 1, 'layer': ['_BatchNorm', 'GroupNorm']}]
  (img_neck): FPN(
    (lateral_convs): ModuleList(
      (0): ConvModule(
        (conv): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1))
      )
    )
    (fpn_convs): ModuleList(
      (0): ConvModule(
        (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
    )
  )
  init_cfg={'type': 'Xavier', 'layer': 'Conv2d', 'distribution': 'uniform'}
  (grid_mask): GridMask()
)
2026/03/14 22:31:07 - bevformer - INFO - Wrapping model
2026/03/14 22:31:07 - bevformer - INFO - Distributed training is not used, all SyncBatchNorm (SyncBN) layers in the model will be automatically reverted to BatchNormXd layers if they are used.
2026/03/14 22:31:07 - bevformer - INFO - Hooks will be executed in the following order:
before_run:
(VERY_HIGH   ) RuntimeInfoHook                    
(BELOW_NORMAL) LoggerHook                         
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointUploader                 
(VERY_LOW    ) CheckpointResumer                  
 -------------------- 
before_train:
(VERY_HIGH   ) RuntimeInfoHook                    
(NORMAL      ) IterTimerHook                      
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointUploader                 
(VERY_LOW    ) CheckpointResumer                  
 -------------------- 
before_train_epoch:
(VERY_HIGH   ) RuntimeInfoHook                    
(NORMAL      ) IterTimerHook                      
(NORMAL      ) DistSamplerSeedHook                
 -------------------- 
before_train_iter:
(VERY_HIGH   ) RuntimeInfoHook                    
(NORMAL      ) IterTimerHook                      
 -------------------- 
after_train_iter:
(VERY_HIGH   ) RuntimeInfoHook                    
(NORMAL      ) IterTimerHook                      
(BELOW_NORMAL) LoggerHook                         
(LOW         ) ParamSchedulerHook                 
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointUploader                 
 -------------------- 
after_train_epoch:
(NORMAL      ) IterTimerHook                      
(LOW         ) ParamSchedulerHook                 
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointUploader                 
 -------------------- 
before_val:
(VERY_HIGH   ) RuntimeInfoHook                    
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointResumer                  
 -------------------- 
before_val_epoch:
(NORMAL      ) IterTimerHook                      
 -------------------- 
before_val_iter:
(NORMAL      ) IterTimerHook                      
 -------------------- 
after_val_iter:
(NORMAL      ) IterTimerHook                      
(BELOW_NORMAL) LoggerHook                         
 -------------------- 
after_val_epoch:
(VERY_HIGH   ) RuntimeInfoHook                    
(NORMAL      ) IterTimerHook                      
(BELOW_NORMAL) LoggerHook                         
(LOW         ) ParamSchedulerHook                 
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointUploader                 
 -------------------- 
after_val:
(VERY_HIGH   ) RuntimeInfoHook                    
 -------------------- 
after_train:
(VERY_HIGH   ) RuntimeInfoHook                    
(VERY_LOW    ) CheckpointHookV2                   
(VERY_LOW    ) CheckpointUploader                 
 -------------------- 
before_test:
(VERY_HIGH   ) RuntimeInfoHook                    
 -------------------- 
before_test_epoch:
(NORMAL      ) IterTimerHook                      
 -------------------- 
before_test_iter:
(NORMAL      ) IterTimerHook                      
 -------------------- 
after_test_iter:
(NORMAL      ) IterTimerHook                      
(BELOW_NORMAL) LoggerHook                         
 -------------------- 
after_test_epoch:
(VERY_HIGH   ) RuntimeInfoHook                    
(NORMAL      ) IterTimerHook                      
(BELOW_NORMAL) LoggerHook                         
 -------------------- 
after_test:
(VERY_HIGH   ) RuntimeInfoHook                    
 -------------------- 
after_run:
(BELOW_NORMAL) LoggerHook                         
 -------------------- 
2026/03/14 22:31:08 - bevformer - INFO - Config:
_dim_ = 256
_ffn_dim_ = 512
_num_levels_ = 1
_pos_dim_ = 128
auto_scale_lr = dict(base_batch_size=16, enable=False)
bev_h_ = 50
bev_w_ = 50
by_epoch = False
class_names = [
    'car',
    'truck',
    'construction_vehicle',
    'bus',
    'trailer',
    'barrier',
    'motorcycle',
    'bicycle',
    'pedestrian',
    'traffic_cone',
]
custom_hooks = [
    dict(
        by_epoch=False,
        clean_local=False,
        interval=5,
        repo_id='5421Project',
        type='CheckpointUploader'),
    dict(repo_id='5421Project', resume_type='last', type='CheckpointResumer'),
]
data = dict(
    nonshuffler_sampler=dict(type='DistributedSampler'),
    samples_per_gpu=1,
    shuffler_sampler=dict(type='DistributedGroupSampler'),
    test=dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_val.pkl',
        bev_size=(
            50,
            50,
        ),
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        frame=[
            -3,
            -2,
            -1,
        ],
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        pipeline=[
            dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
            dict(
                mean=[
                    123.675,
                    116.28,
                    103.53,
                ],
                std=[
                    58.395,
                    57.12,
                    57.375,
                ],
                to_rgb=True,
                type='NormalizeMultiviewImage'),
            dict(
                flip=False,
                img_scale=(
                    800,
                    450,
                ),
                pts_scale_ratio=[
                    1.0,
                ],
                transforms=[
                    dict(
                        scales=[
                            0.5,
                        ], type='RandomScaleImageMultiViewImage'),
                    dict(size_divisor=32, type='PadMultiViewImage'),
                    dict(
                        class_names=[
                            'car',
                            'truck',
                            'construction_vehicle',
                            'bus',
                            'trailer',
                            'barrier',
                            'motorcycle',
                            'bicycle',
                            'pedestrian',
                            'traffic_cone',
                        ],
                        type='CustomDefaultFormatBundle3D'),
                    dict(keys=[
                        'img',
                    ], type='CustomCollect3D'),
                ],
                type='MultiScaleFlipAug3D'),
        ],
        test_mode=True,
        type='CustomNuScenesDataset'),
    train=dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_train.pkl',
        bev_size=(
            50,
            50,
        ),
        box_type_3d='LiDAR',
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        pipeline=[
            dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
            dict(
                type='LoadAnnotations3D',
                with_bbox_3d=True,
                with_label_3d=True),
            dict(
                point_cloud_range=[
                    -51.2,
                    -51.2,
                    -5.0,
                    51.2,
                    51.2,
                    3.0,
                ],
                type='ObjectRangeFilter'),
            dict(
                classes=[
                    'car',
                    'truck',
                    'construction_vehicle',
                    'bus',
                    'trailer',
                    'barrier',
                    'motorcycle',
                    'bicycle',
                    'pedestrian',
                    'traffic_cone',
                ],
                type='ObjectNameFilter'),
            dict(type='PhotoMetricDistortionMultiViewImage'),
            dict(
                mean=[
                    123.675,
                    116.28,
                    103.53,
                ],
                std=[
                    58.395,
                    57.12,
                    57.375,
                ],
                to_rgb=True,
                type='NormalizeMultiviewImage'),
            dict(scales=[
                0.5,
            ], type='RandomScaleImageMultiViewImage'),
            dict(size_divisor=32, type='PadMultiViewImage'),
            dict(
                class_names=[
                    'car',
                    'truck',
                    'construction_vehicle',
                    'bus',
                    'trailer',
                    'barrier',
                    'motorcycle',
                    'bicycle',
                    'pedestrian',
                    'traffic_cone',
                ],
                type='CustomDefaultFormatBundle3D'),
            dict(
                keys=[
                    'gt_bboxes_3d',
                    'gt_labels_3d',
                    'img',
                ],
                type='CustomCollect3D'),
            dict(type='TypeConverter'),
        ],
        queue_length=4,
        test_mode=False,
        type='CustomNuScenesDataset',
        use_valid_flag=True),
    val=dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_val.pkl',
        bev_size=(
            50,
            50,
        ),
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        frame=(),
        frames=[
            -3,
            -2,
            -1,
        ],
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        pipeline=[
            dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
            dict(
                mean=[
                    123.675,
                    116.28,
                    103.53,
                ],
                std=[
                    58.395,
                    57.12,
                    57.375,
                ],
                to_rgb=True,
                type='NormalizeMultiviewImage'),
            dict(
                flip=False,
                img_scale=(
                    800,
                    450,
                ),
                pts_scale_ratio=[
                    1.0,
                ],
                transforms=[
                    dict(
                        scales=[
                            0.5,
                        ], type='RandomScaleImageMultiViewImage'),
                    dict(size_divisor=32, type='PadMultiViewImage'),
                    dict(
                        class_names=[
                            'car',
                            'truck',
                            'construction_vehicle',
                            'bus',
                            'trailer',
                            'barrier',
                            'motorcycle',
                            'bicycle',
                            'pedestrian',
                            'traffic_cone',
                        ],
                        type='CustomDefaultFormatBundle3D'),
                    dict(keys=[
                        'img',
                    ], type='CustomCollect3D'),
                ],
                type='MultiScaleFlipAug3D'),
        ],
        samples_per_gpu=1,
        test_mode=True,
        type='CustomNuScenesDataset'),
    workers_per_gpu=4)
data_root = 'data/nuscenes/v1.0-mini/'
dataset_type = 'CustomNuScenesDataset'
decoder = dict(
    num_layers=6,
    return_intermediate=True,
    transformerlayers=dict(
        attn_cfgs=[
            dict(
                dropout=0.1,
                embed_dims=256,
                num_heads=8,
                type='MultiheadAttention'),
            dict(
                embed_dims=256,
                num_levels=1,
                type='CustomMSDeformableAttention'),
        ],
        ffn_cfgs=dict(
            feedforward_channels=512, ffn_drop=0.1, num_fcs=2, type='FFN'),
        operation_order=(
            'self_attn',
            'norm',
            'cross_attn',
            'norm',
            'ffn',
            'norm',
        ),
        type='DetrTransformerDecoderLayer'),
    type='DetectionTransformerDecoder')
default_hooks = dict(
    checkpoint=dict(
        by_epoch=False,
        interval=5,
        max_keep_ckpts=1,
        save_best=[
            'loss',
            'mAP',
            'NDS',
        ],
        type='CheckpointHookV2'),
    logger=dict(
        interval=2,
        interval_exp_name=1000,
        log_metric_by_epoch=False,
        type='LoggerHook'),
    param_scheduler=dict(type='ParamSchedulerHook'),
    runtime_info=dict(type='RuntimeInfoHook'),
    sampler_seed=dict(type='DistSamplerSeedHook'),
    timer=dict(type='IterTimerHook'))
encoder = dict(
    num_layers=3,
    num_points_in_pillar=8,
    pc_range=[
        -51.2,
        -51.2,
        -5.0,
        51.2,
        51.2,
        3.0,
    ],
    return_intermediate=False,
    transformerlayers=dict(
        attn_cfgs=[
            dict(embed_dims=256, num_levels=1, type='TemporalSelfAttention'),
            dict(
                deformable_attention=dict(
                    embed_dims=256,
                    num_levels=1,
                    num_points=8,
                    type='MSDeformableAttention3D'),
                embed_dims=256,
                pc_range=[
                    -51.2,
                    -51.2,
                    -5.0,
                    51.2,
                    51.2,
                    3.0,
                ],
                type='SpatialCrossAttention'),
        ],
        ffn_cfgs=dict(
            feedforward_channels=512, ffn_drop=0.1, num_fcs=2, type='FFN'),
        operation_order=(
            'self_attn',
            'norm',
            'cross_attn',
            'norm',
            'ffn',
            'norm',
        ),
        type='BEVFormerLayer'),
    type='BEVFormerEncoder')
env_cfg = dict(dist_cfg=dict(backend='nccl'))
experiment_name = 'debug'
file_client_args = dict(backend='disk')
frames = [
    -3,
    -2,
    -1,
]
gpu_ids = range(0, 1)
img_norm_cfg = dict(
    mean=[
        123.675,
        116.28,
        103.53,
    ],
    std=[
        58.395,
        57.12,
        57.375,
    ],
    to_rgb=True)
input_modality = dict(
    use_camera=True,
    use_external=False,
    use_lidar=False,
    use_map=False,
    use_radar=False)
interval = 5
launcher = 'none'
load_from = None
log_interval = 2
log_processor = dict(window_size=20)
lr_config = dict(
    min_lr_ratio=0.001,
    policy='CosineAnnealing',
    warmup='linear',
    warmup_iters=500,
    warmup_ratio=0.3333333333333333)
max_epochs = 5
max_iters = 10
model = dict(
    img_backbone=dict(
        depth=50,
        frozen_stages=1,
        norm_cfg=dict(requires_grad=False, type='BN'),
        norm_eval=True,
        num_stages=4,
        out_indices=(3, ),
        style='pytorch',
        type='ResNet'),
    img_neck=dict(
        add_extra_convs='on_output',
        in_channels=[
            2048,
        ],
        num_outs=1,
        out_channels=256,
        relu_before_extra_convs=True,
        start_level=0,
        type='FPN'),
    pretrained=dict(img='torchvision://resnet50'),
    pts_bbox_head=dict(
        as_two_stage=False,
        bbox_coder=dict(
            max_num=300,
            num_classes=10,
            pc_range=[
                -51.2,
                -51.2,
                -5.0,
                51.2,
                51.2,
                3.0,
            ],
            post_center_range=[
                -61.2,
                -61.2,
                -10.0,
                61.2,
                61.2,
                10.0,
            ],
            type='NMSFreeCoder',
            voxel_size=[
                0.2,
                0.2,
                8,
            ]),
        bev_h=50,
        bev_w=50,
        in_channels=256,
        loss_bbox=dict(loss_weight=0.5, type='L1Loss'),
        loss_cls=dict(
            alpha=0.25,
            gamma=2.0,
            loss_weight=2.0,
            type='FocalLoss',
            use_sigmoid=True),
        loss_iou=dict(loss_weight=0.25, type='GIoULoss'),
        num_classes=10,
        num_query=900,
        positional_encoding=dict(
            col_num_embed=50,
            num_feats=128,
            row_num_embed=50,
            type='LearnedPositionalEncoding'),
        sync_cls_avg_factor=True,
        transformer=dict(
            decoder=dict(
                num_layers=6,
                return_intermediate=True,
                transformerlayers=dict(
                    attn_cfgs=[
                        dict(
                            dropout=0.1,
                            embed_dims=256,
                            num_heads=8,
                            type='MultiheadAttention'),
                        dict(
                            embed_dims=256,
                            num_levels=1,
                            type='CustomMSDeformableAttention'),
                    ],
                    ffn_cfgs=dict(
                        feedforward_channels=512,
                        ffn_drop=0.1,
                        num_fcs=2,
                        type='FFN'),
                    operation_order=(
                        'self_attn',
                        'norm',
                        'cross_attn',
                        'norm',
                        'ffn',
                        'norm',
                    ),
                    type='DetrTransformerDecoderLayer'),
                type='DetectionTransformerDecoder'),
            embed_dims=256,
            encoder=dict(
                num_layers=3,
                num_points_in_pillar=8,
                pc_range=[
                    -51.2,
                    -51.2,
                    -5.0,
                    51.2,
                    51.2,
                    3.0,
                ],
                return_intermediate=False,
                transformerlayers=dict(
                    attn_cfgs=[
                        dict(
                            embed_dims=256,
                            num_levels=1,
                            type='TemporalSelfAttention'),
                        dict(
                            deformable_attention=dict(
                                embed_dims=256,
                                num_levels=1,
                                num_points=8,
                                type='MSDeformableAttention3D'),
                            embed_dims=256,
                            pc_range=[
                                -51.2,
                                -51.2,
                                -5.0,
                                51.2,
                                51.2,
                                3.0,
                            ],
                            type='SpatialCrossAttention'),
                    ],
                    ffn_cfgs=dict(
                        feedforward_channels=512,
                        ffn_drop=0.1,
                        num_fcs=2,
                        type='FFN'),
                    operation_order=(
                        'self_attn',
                        'norm',
                        'cross_attn',
                        'norm',
                        'ffn',
                        'norm',
                    ),
                    type='BEVFormerLayer'),
                type='BEVFormerEncoder'),
            num_cams=6,
            num_feature_levels=1,
            rotate_prev_bev=True,
            type='PerceptionTransformer',
            use_can_bus=True,
            use_shift=True),
        type='BEVFormerHead',
        with_box_refine=True),
    train_cfg=dict(
        pts=dict(
            assigner=dict(
                cls_cost=dict(type='FocalCost', weight=2.0),
                iou_cost=dict(type='SmoothL1Cost', weight=0.25),
                pc_range=[
                    -51.2,
                    -51.2,
                    -5.0,
                    51.2,
                    51.2,
                    3.0,
                ],
                reg_cost=dict(type='BBox3DL1Cost', weight=0.25),
                type='HungarianAssigner3D'),
            grid_size=[
                512,
                512,
                1,
            ],
            out_size_factor=4,
            point_cloud_range=[
                -51.2,
                -51.2,
                -5.0,
                51.2,
                51.2,
                3.0,
            ],
            voxel_size=[
                0.2,
                0.2,
                8,
            ])),
    type='BEVFormerDetector',
    use_grid_mask=True,
    video_test_mode=True)
optim_wrapper = dict(
    optimizer=dict(lr=0.0001, type='AdamW', weight_decay=0.01),
    type='OptimWrapper')
optimizer = dict(lr=0.0001, type='AdamW', weight_decay=0.01)
param_scheduler = dict(
    milestones=[
        1,
        2,
    ], type='MultiStepLR')
point_cloud_range = [
    -51.2,
    -51.2,
    -5.0,
    51.2,
    51.2,
    3.0,
]
pts_bbox_head = dict(
    as_two_stage=False,
    bbox_coder=dict(
        max_num=300,
        num_classes=10,
        pc_range=[
            -51.2,
            -51.2,
            -5.0,
            51.2,
            51.2,
            3.0,
        ],
        post_center_range=[
            -61.2,
            -61.2,
            -10.0,
            61.2,
            61.2,
            10.0,
        ],
        type='NMSFreeCoder',
        voxel_size=[
            0.2,
            0.2,
            8,
        ]),
    bev_h=50,
    bev_w=50,
    in_channels=256,
    loss_bbox=dict(loss_weight=0.5, type='L1Loss'),
    loss_cls=dict(
        alpha=0.25,
        gamma=2.0,
        loss_weight=2.0,
        type='FocalLoss',
        use_sigmoid=True),
    loss_iou=dict(loss_weight=0.25, type='GIoULoss'),
    num_classes=10,
    num_query=900,
    positional_encoding=dict(
        col_num_embed=50,
        num_feats=128,
        row_num_embed=50,
        type='LearnedPositionalEncoding'),
    sync_cls_avg_factor=True,
    transformer=dict(
        decoder=dict(
            num_layers=6,
            return_intermediate=True,
            transformerlayers=dict(
                attn_cfgs=[
                    dict(
                        dropout=0.1,
                        embed_dims=256,
                        num_heads=8,
                        type='MultiheadAttention'),
                    dict(
                        embed_dims=256,
                        num_levels=1,
                        type='CustomMSDeformableAttention'),
                ],
                ffn_cfgs=dict(
                    feedforward_channels=512,
                    ffn_drop=0.1,
                    num_fcs=2,
                    type='FFN'),
                operation_order=(
                    'self_attn',
                    'norm',
                    'cross_attn',
                    'norm',
                    'ffn',
                    'norm',
                ),
                type='DetrTransformerDecoderLayer'),
            type='DetectionTransformerDecoder'),
        embed_dims=256,
        encoder=dict(
            num_layers=3,
            num_points_in_pillar=8,
            pc_range=[
                -51.2,
                -51.2,
                -5.0,
                51.2,
                51.2,
                3.0,
            ],
            return_intermediate=False,
            transformerlayers=dict(
                attn_cfgs=[
                    dict(
                        embed_dims=256,
                        num_levels=1,
                        type='TemporalSelfAttention'),
                    dict(
                        deformable_attention=dict(
                            embed_dims=256,
                            num_levels=1,
                            num_points=8,
                            type='MSDeformableAttention3D'),
                        embed_dims=256,
                        pc_range=[
                            -51.2,
                            -51.2,
                            -5.0,
                            51.2,
                            51.2,
                            3.0,
                        ],
                        type='SpatialCrossAttention'),
                ],
                ffn_cfgs=dict(
                    feedforward_channels=512,
                    ffn_drop=0.1,
                    num_fcs=2,
                    type='FFN'),
                operation_order=(
                    'self_attn',
                    'norm',
                    'cross_attn',
                    'norm',
                    'ffn',
                    'norm',
                ),
                type='BEVFormerLayer'),
            type='BEVFormerEncoder'),
        num_cams=6,
        num_feature_levels=1,
        rotate_prev_bev=True,
        type='PerceptionTransformer',
        use_can_bus=True,
        use_shift=True),
    type='BEVFormerHead',
    with_box_refine=True)
queue_length = 4
resume = False
scales = [
    0.5,
]
test_cfg = dict(max_iters=1)
test_dataloader = dict(
    batch_size=1,
    collate_fn=dict(type='test_collate'),
    dataset=dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_val.pkl',
        bev_size=(
            50,
            50,
        ),
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        frame=[
            -3,
            -2,
            -1,
        ],
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        pipeline=[
            dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
            dict(
                mean=[
                    123.675,
                    116.28,
                    103.53,
                ],
                std=[
                    58.395,
                    57.12,
                    57.375,
                ],
                to_rgb=True,
                type='NormalizeMultiviewImage'),
            dict(
                flip=False,
                img_scale=(
                    800,
                    450,
                ),
                pts_scale_ratio=[
                    1.0,
                ],
                transforms=[
                    dict(
                        scales=[
                            0.5,
                        ], type='RandomScaleImageMultiViewImage'),
                    dict(size_divisor=32, type='PadMultiViewImage'),
                    dict(
                        class_names=[
                            'car',
                            'truck',
                            'construction_vehicle',
                            'bus',
                            'trailer',
                            'barrier',
                            'motorcycle',
                            'bicycle',
                            'pedestrian',
                            'traffic_cone',
                        ],
                        type='CustomDefaultFormatBundle3D'),
                    dict(keys=[
                        'img',
                    ], type='CustomCollect3D'),
                ],
                type='MultiScaleFlipAug3D'),
        ],
        test_mode=True,
        type='CustomNuScenesDataset'),
    num_workers=0,
    sampler=dict(shuffle=True, type='DefaultSampler'))
test_evaluator = dict(metrics=[
    dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_val.pkl',
        data_root='data/nuscenes/v1.0-mini/',
        type='src.NuScenesMetric',
        version='v1.0-mini'),
])
test_max_iters = 1
test_pipeline = [
    dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
    dict(
        mean=[
            123.675,
            116.28,
            103.53,
        ],
        std=[
            58.395,
            57.12,
            57.375,
        ],
        to_rgb=True,
        type='NormalizeMultiviewImage'),
    dict(
        flip=False,
        img_scale=(
            800,
            450,
        ),
        pts_scale_ratio=[
            1.0,
        ],
        transforms=[
            dict(scales=[
                0.5,
            ], type='RandomScaleImageMultiViewImage'),
            dict(size_divisor=32, type='PadMultiViewImage'),
            dict(
                class_names=[
                    'car',
                    'truck',
                    'construction_vehicle',
                    'bus',
                    'trailer',
                    'barrier',
                    'motorcycle',
                    'bicycle',
                    'pedestrian',
                    'traffic_cone',
                ],
                type='CustomDefaultFormatBundle3D'),
            dict(keys=[
                'img',
            ], type='CustomCollect3D'),
        ],
        type='MultiScaleFlipAug3D'),
]
train_cfg = dict(by_epoch=False, max_epochs=5, max_iters=10, val_interval=5)
train_dataloader = dict(
    batch_size=1,
    collate_fn=dict(type='train_collate'),
    dataset=dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_train.pkl',
        bev_size=(
            50,
            50,
        ),
        box_type_3d='LiDAR',
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        pipeline=[
            dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
            dict(
                type='LoadAnnotations3D',
                with_bbox_3d=True,
                with_label_3d=True),
            dict(
                point_cloud_range=[
                    -51.2,
                    -51.2,
                    -5.0,
                    51.2,
                    51.2,
                    3.0,
                ],
                type='ObjectRangeFilter'),
            dict(
                classes=[
                    'car',
                    'truck',
                    'construction_vehicle',
                    'bus',
                    'trailer',
                    'barrier',
                    'motorcycle',
                    'bicycle',
                    'pedestrian',
                    'traffic_cone',
                ],
                type='ObjectNameFilter'),
            dict(type='PhotoMetricDistortionMultiViewImage'),
            dict(
                mean=[
                    123.675,
                    116.28,
                    103.53,
                ],
                std=[
                    58.395,
                    57.12,
                    57.375,
                ],
                to_rgb=True,
                type='NormalizeMultiviewImage'),
            dict(scales=[
                0.5,
            ], type='RandomScaleImageMultiViewImage'),
            dict(size_divisor=32, type='PadMultiViewImage'),
            dict(
                class_names=[
                    'car',
                    'truck',
                    'construction_vehicle',
                    'bus',
                    'trailer',
                    'barrier',
                    'motorcycle',
                    'bicycle',
                    'pedestrian',
                    'traffic_cone',
                ],
                type='CustomDefaultFormatBundle3D'),
            dict(
                keys=[
                    'gt_bboxes_3d',
                    'gt_labels_3d',
                    'img',
                ],
                type='CustomCollect3D'),
            dict(type='TypeConverter'),
        ],
        queue_length=4,
        test_mode=False,
        type='CustomNuScenesDataset',
        use_valid_flag=True),
    num_workers=0,
    sampler=dict(shuffle=True, type='DefaultSampler'))
train_pipeline = [
    dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
    dict(
        point_cloud_range=[
            -51.2,
            -51.2,
            -5.0,
            51.2,
            51.2,
            3.0,
        ],
        type='ObjectRangeFilter'),
    dict(
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        type='ObjectNameFilter'),
    dict(type='PhotoMetricDistortionMultiViewImage'),
    dict(
        mean=[
            123.675,
            116.28,
            103.53,
        ],
        std=[
            58.395,
            57.12,
            57.375,
        ],
        to_rgb=True,
        type='NormalizeMultiviewImage'),
    dict(scales=[
        0.5,
    ], type='RandomScaleImageMultiViewImage'),
    dict(size_divisor=32, type='PadMultiViewImage'),
    dict(
        class_names=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        type='CustomDefaultFormatBundle3D'),
    dict(
        keys=[
            'gt_bboxes_3d',
            'gt_labels_3d',
            'img',
        ], type='CustomCollect3D'),
    dict(type='TypeConverter'),
]
transformer = dict(
    decoder=dict(
        num_layers=6,
        return_intermediate=True,
        transformerlayers=dict(
            attn_cfgs=[
                dict(
                    dropout=0.1,
                    embed_dims=256,
                    num_heads=8,
                    type='MultiheadAttention'),
                dict(
                    embed_dims=256,
                    num_levels=1,
                    type='CustomMSDeformableAttention'),
            ],
            ffn_cfgs=dict(
                feedforward_channels=512, ffn_drop=0.1, num_fcs=2, type='FFN'),
            operation_order=(
                'self_attn',
                'norm',
                'cross_attn',
                'norm',
                'ffn',
                'norm',
            ),
            type='DetrTransformerDecoderLayer'),
        type='DetectionTransformerDecoder'),
    embed_dims=256,
    encoder=dict(
        num_layers=3,
        num_points_in_pillar=8,
        pc_range=[
            -51.2,
            -51.2,
            -5.0,
            51.2,
            51.2,
            3.0,
        ],
        return_intermediate=False,
        transformerlayers=dict(
            attn_cfgs=[
                dict(
                    embed_dims=256, num_levels=1,
                    type='TemporalSelfAttention'),
                dict(
                    deformable_attention=dict(
                        embed_dims=256,
                        num_levels=1,
                        num_points=8,
                        type='MSDeformableAttention3D'),
                    embed_dims=256,
                    pc_range=[
                        -51.2,
                        -51.2,
                        -5.0,
                        51.2,
                        51.2,
                        3.0,
                    ],
                    type='SpatialCrossAttention'),
            ],
            ffn_cfgs=dict(
                feedforward_channels=512, ffn_drop=0.1, num_fcs=2, type='FFN'),
            operation_order=(
                'self_attn',
                'norm',
                'cross_attn',
                'norm',
                'ffn',
                'norm',
            ),
            type='BEVFormerLayer'),
        type='BEVFormerEncoder'),
    num_cams=6,
    num_feature_levels=1,
    rotate_prev_bev=True,
    type='PerceptionTransformer',
    use_can_bus=True,
    use_shift=True)
val_cfg = dict(max_iters=1)
val_dataloader = dict(
    batch_size=1,
    collate_fn=dict(type='test_collate'),
    dataset=dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_val.pkl',
        bev_size=(
            50,
            50,
        ),
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        frame=(),
        frames=[
            -3,
            -2,
            -1,
        ],
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        pipeline=[
            dict(to_float32=True, type='LoadMultiViewImageFromFiles'),
            dict(
                mean=[
                    123.675,
                    116.28,
                    103.53,
                ],
                std=[
                    58.395,
                    57.12,
                    57.375,
                ],
                to_rgb=True,
                type='NormalizeMultiviewImage'),
            dict(
                flip=False,
                img_scale=(
                    800,
                    450,
                ),
                pts_scale_ratio=[
                    1.0,
                ],
                transforms=[
                    dict(
                        scales=[
                            0.5,
                        ], type='RandomScaleImageMultiViewImage'),
                    dict(size_divisor=32, type='PadMultiViewImage'),
                    dict(
                        class_names=[
                            'car',
                            'truck',
                            'construction_vehicle',
                            'bus',
                            'trailer',
                            'barrier',
                            'motorcycle',
                            'bicycle',
                            'pedestrian',
                            'traffic_cone',
                        ],
                        type='CustomDefaultFormatBundle3D'),
                    dict(keys=[
                        'img',
                    ], type='CustomCollect3D'),
                ],
                type='MultiScaleFlipAug3D'),
        ],
        samples_per_gpu=1,
        test_mode=True,
        type='CustomNuScenesDataset'),
    num_workers=0,
    sampler=dict(shuffle=True, type='DefaultSampler'))
val_evaluator = dict(metrics=[
    dict(
        ann_file='data/nuscenes/v1.0-mini/nuscenes_infos_temporal_val.pkl',
        classes=[
            'car',
            'truck',
            'construction_vehicle',
            'bus',
            'trailer',
            'barrier',
            'motorcycle',
            'bicycle',
            'pedestrian',
            'traffic_cone',
        ],
        data_root='data/nuscenes/v1.0-mini/',
        jsonfile_prefix='results',
        modality=dict(
            use_camera=True,
            use_external=False,
            use_lidar=False,
            use_map=False,
            use_radar=False),
        plot_every_run=True,
        plot_examples=1,
        type='src.NuScenesMetric',
        version='v1.0-mini'),
])
val_interval = 5
val_max_iters = 1
version = 'v1.0-mini'
visualizer = dict(
    type='Visualizer',
    vis_backends=[
        dict(type='LocalVisBackend'),
        dict(type='TensorboardVisBackend'),
    ])
voxel_size = [
    0.2,
    0.2,
    8,
]
work_dir = 'experiment'

2026/03/14 22:31:08 - bevformer - INFO - See full config in 'experiment/debug/bevformer_tiny_test.py'.
2026/03/14 22:31:10 - bevformer - INFO - Checkpoints will be saved to 'experiment/debug' after every 5 steps.
2026/03/14 22:31:10 - bevformer - INFO - Initialize best checkpoints by train phase.
2026/03/14 22:31:10 - bevformer - INFO - Set best path for 'loss' None.
2026/03/14 22:31:10 - bevformer - INFO - Set best path for 'mAP' None.
2026/03/14 22:31:10 - bevformer - INFO - Set best path for 'NDS' None.
2026/03/14 22:31:10 - bevformer - INFO - The best checkpoints will be saved to 'experiment/debug' based on ['loss', 'mAP', 'NDS'] with rules ['less', 'greater', 'greater'] after every 5 steps.
2026/03/14 22:31:10 - bevformer - INFO - Keep maximum 1 checkpoints in local.
2026/03/14 22:31:13 - bevformer - INFO - Created '5421Project/debug' to save checkpoints (see 'https://huggingface.co/5421Project/debug').
2026/03/14 22:31:13 - bevformer - INFO - Checkpoints will be pushed to repo 'https://huggingface.co/5421Project/debug' after every 5 steps.
2026/03/14 22:31:55 - bevformer - INFO - Epoch(train) [1][  2/323]  lr: 1.0000e-04  eta: 0:02:49  time: 21.1329  data_time: 1.0395  loss: 54.7820  loss_cls: 2.2636  loss_bbox: 7.1453  d0.loss_cls: 2.2556  d0.loss_bbox: 6.7686  d1.loss_cls: 2.2362  d1.loss_bbox: 6.8347  d2.loss_cls: 2.1509  d2.loss_bbox: 6.7891  d3.loss_cls: 2.2262  d3.loss_bbox: 6.8406  d4.loss_cls: 2.2757  d4.loss_bbox: 6.9953
2026/03/14 22:32:35 - bevformer - INFO - Epoch(train) [1][  4/323]  lr: 1.0000e-04  eta: 0:02:03  time: 20.5688  data_time: 1.0071  loss: 52.4912  loss_cls: 2.2311  loss_bbox: 6.6259  d0.loss_cls: 2.2306  d0.loss_bbox: 6.5213  d1.loss_cls: 2.2324  d1.loss_bbox: 6.5294  d2.loss_cls: 2.1107  d2.loss_bbox: 6.5639  d3.loss_cls: 2.1472  d3.loss_bbox: 6.5297  d4.loss_cls: 2.2651  d4.loss_bbox: 6.5040
2026/03/14 22:32:56 - bevformer - INFO - Saving checkpoint at 5 iterations
2026/03/14 22:32:57 - bevformer - INFO - Saving best checkpoints...
2026/03/14 22:32:57 - bevformer - INFO - Set 'best_score_loss' to +/-inf as it is not in message hub.
2026/03/14 22:32:57 - bevformer - INFO - [loss]: Best score: inf, current score: 43.642234802246094
2026/03/14 22:32:58 - bevformer - INFO - The best checkpoint with 43.6422 loss at 5 iter is saved to 'best_loss_iter_5.pth'.
2026/03/14 22:32:58 - bevformer - INFO - Pushing checkpoint at 5 steps...
2026/03/14 22:34:31 - bevformer - INFO - Pushed last checkpoint 'experiment/debug/iter_5.pth' to repo
2026/03/14 22:34:35 - bevformer - INFO - Pushed best checkpoint 'best_loss_iter_5.pth' of [loss]...
2026/03/14 22:34:49 - bevformer - INFO - Epoch(val) [0][81/81]    NDS: 0.0000  mAP: 0.0000  data_time: 0.8768  time: 2.6421
2026/03/14 22:34:49 - bevformer - INFO - Save best checkpoints after val epoch.
2026/03/14 22:34:49 - bevformer - INFO - Saving best checkpoints...
2026/03/14 22:34:49 - bevformer - INFO - Set 'best_score_mAP' to +/-inf as it is not in message hub.
2026/03/14 22:34:49 - bevformer - INFO - [mAP]: Best score: -inf, current score: 3.561391434020159e-05
2026/03/14 22:34:50 - bevformer - INFO - The best checkpoint with 0.0000 mAP at 5 iter is saved to 'best_mAP_iter_5.pth'.
2026/03/14 22:34:50 - bevformer - INFO - Set 'best_score_NDS' to +/-inf as it is not in message hub.
2026/03/14 22:34:50 - bevformer - INFO - [NDS]: Best score: -inf, current score: 1.7806957170100794e-05
2026/03/14 22:34:51 - bevformer - INFO - The best checkpoint with 0.0000 NDS at 5 iter is saved to 'best_NDS_iter_5.pth'.
2026/03/14 22:34:51 - bevformer - INFO - Resaving checkpoint at 5 iter...
2026/03/14 22:34:55 - bevformer - INFO - Pushed best checkpoint 'best_mAP_iter_5.pth' of [mAP]...
2026/03/14 22:34:57 - bevformer - INFO - Pushed best checkpoint 'best_NDS_iter_5.pth' of [NDS]...
2026/03/14 22:35:20 - bevformer - INFO - Epoch(train) [1][  6/323]  lr: 1.0000e-04  eta: 0:01:36  time: 24.1120  data_time: 4.3466  loss: 51.6401  loss_cls: 2.1800  loss_bbox: 6.4342  d0.loss_cls: 2.1850  d0.loss_bbox: 6.4393  d1.loss_cls: 2.2004  d1.loss_bbox: 6.4757  d2.loss_cls: 2.0870  d2.loss_bbox: 6.4939  d3.loss_cls: 2.1014  d3.loss_bbox: 6.4365  d4.loss_cls: 2.2068  d4.loss_bbox: 6.3998
2026/03/14 22:36:00 - bevformer - INFO - Epoch(train) [1][  8/323]  lr: 1.0000e-04  eta: 0:00:46  time: 23.1149  data_time: 3.4998  loss: 49.8073  loss_cls: 2.1378  loss_bbox: 6.1821  d0.loss_cls: 2.1482  d0.loss_bbox: 6.1880  d1.loss_cls: 2.1749  d1.loss_bbox: 6.1841  d2.loss_cls: 2.0729  d2.loss_bbox: 6.2197  d3.loss_cls: 2.0748  d3.loss_bbox: 6.1526  d4.loss_cls: 2.1567  d4.loss_bbox: 6.1156
2026/03/14 22:36:42 - bevformer - INFO - Epoch(train) [1][ 10/323]  lr: 1.0000e-04  eta: 0:00:00  time: 22.7019  data_time: 2.9867  loss: 50.4065  loss_cls: 2.1042  loss_bbox: 6.3351  d0.loss_cls: 2.1262  d0.loss_bbox: 6.3189  d1.loss_cls: 2.1572  d1.loss_bbox: 6.2772  d2.loss_cls: 2.0582  d2.loss_bbox: 6.3471  d3.loss_cls: 2.0630  d3.loss_bbox: 6.2461  d4.loss_cls: 2.1094  d4.loss_bbox: 6.2639
2026/03/14 22:36:42 - bevformer - INFO - Saving checkpoint at 10 iterations
2026/03/14 22:36:44 - bevformer - INFO - Saving best checkpoints...
2026/03/14 22:36:44 - bevformer - INFO - Got best score ['loss'] from message hub
2026/03/14 22:36:44 - bevformer - INFO - [loss]: Best score: 43.642234802246094, current score: 50.985355377197266
2026/03/14 22:36:44 - bevformer - INFO - Pushing checkpoint at 10 steps...
2026/03/14 22:37:18 - bevformer - INFO - Pushed last checkpoint 'experiment/debug/iter_10.pth' to repo
2026/03/14 22:37:19 - bevformer - INFO - Removed 'iter_5.pth' from repo
2026/03/14 22:37:32 - bevformer - INFO - Epoch(val) [0][81/81]    NDS: 0.0221  mAP: 0.0000  data_time: 0.7784  time: 2.4680
2026/03/14 22:37:32 - bevformer - INFO - Save best checkpoints after val epoch.
2026/03/14 22:37:32 - bevformer - INFO - Saving best checkpoints...
2026/03/14 22:37:32 - bevformer - INFO - Got best score ['mAP'] from message hub
2026/03/14 22:37:32 - bevformer - INFO - [mAP]: Best score: 3.561391434020159e-05, current score: 7.014590347923641e-07
2026/03/14 22:37:32 - bevformer - INFO - Got best score ['NDS'] from message hub
2026/03/14 22:37:32 - bevformer - INFO - [NDS]: Best score: 1.7806957170100794e-05, current score: 0.02213961767715449
2026/03/14 22:37:32 - bevformer - INFO - The previous best checkpoint 'experiment/debug/best_NDS_iter_5.pth' is removed
2026/03/14 22:37:33 - bevformer - INFO - The best checkpoint with 0.0221 NDS at 10 iter is saved to 'best_NDS_iter_10.pth'.
2026/03/14 22:37:33 - bevformer - INFO - Resaving checkpoint at 10 iter...
2026/03/14 22:37:36 - bevformer - INFO - Pushed best checkpoint 'best_NDS_iter_10.pth' of [NDS]...
2026/03/14 22:37:36 - bevformer - INFO - Removed 'best_NDS_iter_5.pth' from repo
2026/03/14 22:37:37 - bevformer - INFO - Pushing visualizing data and safetensors to repo after training...