In training, the actual GPU memory usage is significantly higher than the model's memory footprint

[nuonepeaceyy]

Prerequisite

• [X] I have searched Issues and Discussions but cannot get the expected help.
• [X] I have read the FAQ documentation but cannot get the expected help.
• [x] The bug has not been fixed in the latest version (dev-1.x) or latest version (dev-1.0).

Task

I'm using the official example scripts/configs for the officially supported tasks/models/datasets.

Branch

main branch https://github.com/open-mmlab/mmdetection3d

Environment

sys.platform: linux Python: 3.8.19 (default, Mar 20 2024, 19:58:24) [GCC 11.2.0] CUDA available: True MUSA available: False numpy_random_seed: 2147483648 GPU 0: NVIDIA GeForce RTX 4090 CUDA_HOME: /usr/local/cuda-11.8 NVCC: Cuda compilation tools, release 11.8, V11.8.89 GCC: gcc (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0 PyTorch: 1.13.1+cu117 PyTorch compiling details: PyTorch built with:

• GCC 9.3
• 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 v2.6.0 (Git Hash 52b5f107dd9cf10910aaa19cb47f3abf9b349815)
• OpenMP 201511 (a.k.a. OpenMP 4.5)
• LAPACK is enabled (usually provided by MKL)
• NNPACK is enabled
• CPU capability usage: AVX2
• CUDA Runtime 11.7
• 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;-gencode;arch=compute_86,code=sm_86
• CuDNN 8.5
• Magma 2.6.1
• Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.7, CUDNN_VERSION=8.5.0, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -fabi-version=11 -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_KINETO -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wunused-local-typedefs -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -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 -Werror=cast-function-type -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.13.1, USE_CUDA=ON, USE_CUDNN=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, USE_ROCM=OFF,

TorchVision: 0.14.1+cu117 OpenCV: 4.9.0 MMEngine: 0.10.3 MMDetection: 3.3.0 MMDetection3D: 1.4.0+ spconv2.0: True

Reproduces the problem - code sample

I didn't make changes to the code.

Reproduces the problem - command or script

bash tools/dist_train.sh projects/BEVFusion/configs/bevfusion_lidar-cam_voxel0075_second_secfpn_8xb4-cyclic-20e_nus-3d.py 1

Reproduces the problem - error message

Thank you for your work. I recently transitioned my work from OpenPCDet to the MMDetection3D framework. However, when I attempted to retrain BEVfusion following the tutorial in the README, I noticed that the actual GPU memory usage is much higher than the memory usage recorded in the logs for the model. In fact, the memory usage recorded in the logs is consistent with the memory usage when training BEVfusion using the OpenPCDet framework. I would like to understand what is causing this additional memory usage and what steps I should take to avoid it. And, the GPU memory usage seems to continuously increase as the training progresses.

Additional information

No response