Issue about the pretrained model

[chaoqunwangcs]

Thanks to the work, I test the pretrained model in issue #19 . Compared with the reported results in Tab.1 in the paper, I got different performance with the uploaded pretrain model of lidar branch, while the camera branch is completely the same. The performance of lidar branch is mAP:64.58, NDS:69.69, and the performance in Tab.1 is mAP: 64.9, NDS: 69.9. Is there any wrong of misunderstanding?

The environment is:
sys.platform: linux
Python: 3.8.3 (default, Jul 2 2020, 16:21:59) [GCC 7.3.0]
CUDA available: True
GPU 0: NVIDIA A100-SXM4-40GB MIG 3g.20gb
CUDA_HOME: /usr/local/cuda
NVCC: Build cuda_11.2.r11.2/compiler.29558016_0
GCC: gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
PyTorch: 1.7.0+cu110
PyTorch compiling details: PyTorch built with:

• GCC 7.3
• C++ Version: 201402
• Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
• Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
• OpenMP 201511 (a.k.a. OpenMP 4.5)
• NNPACK is enabled
• CPU capability usage: AVX2
• CUDA Runtime 11.0
• NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80
• CuDNN 8.0.4
• Magma 2.5.2
• Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON,

TorchVision: 0.8.1+cu110
OpenCV: 4.6.0
MMCV: 1.4.0
MMCV Compiler: GCC 9.3
MMCV CUDA Compiler: 11.2
MMDetection: 2.11.0
MMDetection3D: 0.11.0+fb4384c

[IranQin]

Hello, I got the same problem, the lidar pretrained model could not get the same NDS and mAP as you reported.
Using this lidar pretrained model, I trained fusion model and get 71.56NDS and 68.85mAP, which are much lower than the repo reported. Is this problem caused by the lower performance lidar pretrained model or other reasons.