ZeroDivisionError: float division by zero

[xingkongliang]

Describe the bug

2021-12-29 08:46:30.550 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 5e-324
2021-12-29 08:46:30.551 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 5e-324
2021-12-29 08:46:31.863 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.863 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.864 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.864 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.864 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.864 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.864 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0
2021-12-29 08:46:31.865 Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to 0.0

2021-12-29 08:46:39.965 File "./tools/train.py", line 188, in <module>
2021-12-29 08:46:39.965 main()
2021-12-29 08:46:39.965 File "./tools/train.py", line 177, in main
2021-12-29 08:46:39.965 train_detector(
2021-12-29 08:46:39.965 File "mmdet/apis/train.py", line 186, in train_detector
2021-12-29 08:46:39.965 runner.run(data_loaders, cfg.workflow)
2021-12-29 08:46:39.965 File "/opt/conda/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 127, in run
2021-12-29 08:46:39.965 epoch_runner(data_loaders[i], **kwargs)
2021-12-29 08:46:39.965 File "/opt/conda/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 51, in train
2021-12-29 08:46:39.965 self.call_hook('after_train_iter')
2021-12-29 08:46:39.965 File "/opt/conda/lib/python3.8/site-packages/mmcv/runner/base_runner.py", line 307, in call_hook
2021-12-29 08:46:39.966 getattr(hook, fn_name)(self)
2021-12-29 08:46:39.966 File "mmdet/utils/optimizer.py", line 26, in after_train_iter
2021-12-29 08:46:39.966 scaled_loss.backward()
2021-12-29 08:46:39.966 File "/opt/conda/lib/python3.8/contextlib.py", line 120, in __exit__
2021-12-29 08:46:39.966 next(self.gen)
2021-12-29 08:46:39.966 File "/opt/conda/lib/python3.8/site-packages/apex/amp/handle.py", line 123, in scale_loss
2021-12-29 08:46:39.966 optimizer._post_amp_backward(loss_scaler)
2021-12-29 08:46:39.966 File "/opt/conda/lib/python3.8/site-packages/apex/amp/_process_optimizer.py", line 249, in post_backward_no_master_weights
2021-12-29 08:46:39.966 post_backward_models_are_masters(scaler, params, stashed_grads)
2021-12-29 08:46:39.966 File "/opt/conda/lib/python3.8/site-packages/apex/amp/_process_optimizer.py", line 131, in post_backward_models_are_masters
2021-12-29 08:46:39.966 scaler.unscale_with_stashed(
2021-12-29 08:46:39.966 File "/opt/conda/lib/python3.8/site-packages/apex/amp/scaler.py", line 176, in unscale_with_stashed
2021-12-29 08:46:39.966 out_scale/grads_have_scale, # 1./scale,
2021-12-29 08:46:39.966 ZeroDivisionError: float division by zero

Reproduction

1. What command or script did you run?
   run configs/cbnet/htc_cbv2_swin_base_patch4_window7_mstrain_400-1400_giou_4conv1f_adamw_20e_coco.py

2. Did you make any modifications on the code or config? Did you understand what you have modified?
   no

3. What dataset did you use?
   coco

Environment

2021-12-28 02:35:44,932 - mmdet - INFO - Environment info:
------------------------------------------------------------
sys.platform: linux
Python: 3.8.5 (default, Sep  4 2020, 07:30:14) [GCC 7.3.0]
CUDA available: True
GPU 0,1,2,3,4,5,6,7: Tesla V100-SXM2-32GB
CUDA_HOME: /usr/local/cuda
NVCC: Build cuda_11.0_bu.TC445_37.28845127_0
GCC: gcc (GCC) 8.3.1 20191121 (Red Hat 8.3.1-5)
PyTorch: 1.7.0+cu110
PyTorch compiling details: PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.2 Product Build 20200624 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: 4.5.1
MMCV: 1.3.18
MMCV Compiler: GCC 7.3
MMCV CUDA Compiler: 11.0
MMDetection: 2.14.0+
------------------------------------------------------------