Object detection task

[hbzhou530]

很棒的工作！
我正尝试复现目标检测任务，遇到如下问题，请多多指教！！

Q1. Object detection's pre-training
论文当中描述使用SemanticKITTI预训练，但代码里给出是在nuscenes上（cfgs/pretrain_ns_spconv.yaml）。我试着调整了下，写了SK的配置文件pretrain_sk_spconv.zip。但是多次运行在第二个epoch时，报错：

CUDA runtime error: an illegal memory access was encountered (700) in apply_lu_factor_batched_magma at /opt/conda/conda-bld/pytorch_1656352464346/work/aten/src/ATen/native/cuda/linalg/BatchLinearAlgebra.cpp:1961
CUDA runtime error: an illegal memory access was encountered (700) in magma_queue_destroy_internal at /opt/conda/conda-bld/magma-cuda113_1619629459349/work/interface_cuda/interface.cpp:944
CUDA runtime error: an illegal memory access was encountered (700) in magma_queue_destroy_internal at /opt/conda/conda-bld/magma-cuda113_1619629459349/work/interface_cuda/interface.cpp:945
CUDA runtime error: an illegal memory access was encountered (700) in magma_queue_destroy_internal at /opt/conda/conda-bld/magma-cuda113_1619629459349/work/interface_cuda/interface.cpp:946
epochs:   2%|█▌                                                                         | 1/50 [31:39<25:51:18, 1899.55s/it, loss=15]
Traceback (most recent call last):
  File "train.py", line 300, in <module>
    main(0, 1)
  File "train.py", line 199, in main
    multigpu=multigpu
  File "train.py", line 246, in train
    loss, metrics = model(batch)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
  File "/code/NCLR/bevlab/models.py", line 453, in forward
    x_cam, R, t, err_2d, err_3d = efficient_pnp(keypoints_coordinates, y_uncalibrated)
  File "/code/NCLR/bevlab/pnp.py", line 632, in efficient_pnp
    for c_cam in c_cam_variants
  File "/code/NCLR/bevlab/pnp.py", line 632, in <listcomp>
    for c_cam in c_cam_variants
  File "/code/NCLR/bevlab/pnp.py", line 396, in _compute_norm_sign_scaling_factor
    x_world, x_cam, weight, estimate_scale=True
  File "/code/NCLR/bevlab/pnp.py", line 193, in corresponding_points_alignment
    E[:, -1, -1] = torch.det(R_test)
RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.

但是，ns的也试了试，pretrain_ns_spconv.yaml 运行正常。
想到目标检测下游任务是在kitti上进行，感觉更应该是用SemanticKITTI进行的预训练。如果是，请帮忙检查一下我的配置文件。

Q2. Object detection's downstream
请问怎么进行10%、20%、50% ratio设置。如果方便，请分享下相关代码或配置。感谢！

再次感谢！！