[Bug] RuntimeError: CUDA out of memory. Tried to allocate 1048475.67 GiB

[lin199811]

Prerequisite

• I have searched Issues and Discussions but cannot get the expected help.
• I have read the FAQ documentation but cannot get the expected help.
• 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

System environment:
sys.platform: linux
Python: 3.8.18 (default, Sep 11 2023, 13:40:15) [GCC 11.2.0]
CUDA available: True
MUSA available: False
numpy_random_seed: 1591519926
GPU 0,1,2,3,4,5,6,7: NVIDIA GeForce RTX 3090
CUDA_HOME: /usr
NVCC: Cuda compilation tools, release 10.1, V10.1.24
GCC: gcc (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
PyTorch: 1.11.0
PyTorch compiling details: PyTorch built with:

• GCC 7.3

• C++ Version: 201402

• Intel(R) oneAPI Math Kernel Library Version 2023.1-Product Build 20230303 for Intel(R) 64 architecture applications

• Intel(R) MKL-DNN v2.5.2 (Git Hash a9302535553c73243c632ad3c4c80beec3d19a1e)

• OpenMP 201511 (a.k.a. OpenMP 4.5)

• LAPACK is enabled (usually provided by MKL)

• NNPACK is enabled

• CPU capability usage: AVX2

• CUDA Runtime 11.3

• 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_61,code=sm_61;-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;-gencode;arch=compute_37,code=compute_37

• CuDNN 8.2

• Magma 2.5.2

• Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.3, CUDNN_VERSION=8.2.0, CXX_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/c++, CXX_FLAGS= -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 -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -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, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.11.0, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=OFF, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF,

  TorchVision: 0.12.0
  OpenCV: 4.9.0
  MMEngine: 0.10.3

Runtime environment:
cudnn_benchmark: False
mp_cfg: {'mp_start_method': 'fork', 'opencv_num_threads': 0}
dist_cfg: {'backend': 'nccl'}
seed: 1591519926
Distributed launcher: none
Distributed training: False
GPU number: 1

Reproduces the problem - code sample

Traceback (most recent call last):
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/contextlib.py", line 131, in exit
self.gen.throw(type, value, traceback)
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/mmengine/optim/optimizer/optimizer_wrapper.py", line 283, in optim_context
yield
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 114, in train_step
losses = self._run_forward(data, mode='loss') # type: ignore
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 361, in _run_forward
results = self(**data, mode=mode)
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "/storage1/Fudongyi/AutoDrive/code/EmbodiedScan/embodiedscan/models/detectors/sparse_featfusion_single_stage.py", line 325, in forward
return self.loss(inputs, data_samples, **kwargs)
File "/storage1/Fudongyi/AutoDrive/code/EmbodiedScan/embodiedscan/models/detectors/sparse_featfusion_single_stage.py", line 242, in loss
losses = self.bbox_head.loss(x, batch_data_samples, **kwargs)
File "/storage1/Fudongyi/AutoDrive/code/EmbodiedScan/embodiedscan/models/dense_heads/fcaf3d_head.py", line 1037, in loss
outs = self(x)
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "/storage1/Fudongyi/AutoDrive/code/EmbodiedScan/embodiedscan/models/dense_heads/fcaf3d_head.py", line 1010, in forward
x = self._prune(x, prune_score)
File "/storage1/Fudongyi/AutoDrive/code/EmbodiedScan/embodiedscan/models/dense_heads/fcaf3d_head.py", line 1103, in _prune
interpolated_scores = scores.features_at_coordinates(coordinates)
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/MinkowskiEngine-0.5.4-py3.8-linux-x86_64.egg/MinkowskiEngine/MinkowskiSparseTensor.py", line 713, in features_at_coordinates
return MinkowskiInterpolationFunction().apply(
File "/home/fudongyi/anaconda3/envs/embodiedscan/lib/python3.8/site-packages/MinkowskiEngine-0.5.4-py3.8-linux-x86_64.egg/MinkowskiEngine/MinkowskiInterpolation.py", line 52, in forward
out_feat, in_map, out_map, weights = fw_fn(
RuntimeError: CUDA out of memory. Tried to allocate 1048475.67 GiB (GPU 0; 23.70 GiB total capacity; 1.47 GiB already allocated; 20.20 GiB free; 1.50 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

Reproduces the problem - command or script

python tools/train.py configs/detection/mv-det3d_8xb4_embodiedscan-3d-284class-9dof.py --work-dir=work_dirs/mv-3ddet

Reproduces the problem - error message

There show the fw_fn() function tried to allocate 1048475.67 GiB at GPU. I think it is a bug. how can I solve this problem?

Additional information

No response

[lin199811]

my train_dataloader is set (batch_size=1, num_workers=1)

[Tai-Wang]

Do you use our officially provided data? It is really strange to allocate such huge amount of memory for the multi-view 3D detection model.

[lin199811]

I found this bug in MinkowskiEngine. I modify and rebuid the MinkowskiEngine code ( in src/spmm.cu coo_spmm function, I change the nnz to static_caststd::size_t(nnz). Therefore, this bug never seen again.

[AmingWu]

@lin199811 How long does your method need to train?

[Zhaooyy]

there's a lot of nnz in src/spmm.cu, should I change them all to static_caststd::size_t(nnz) or just some of them?