【算法】CenterNet运行demo时报错：RuntimeError: CUDA error: out of memory

问题描述

python demo.py ctdet --demo /home/vincent/Code/CenterNet-master/images/ --load_model /home/vincent/Code/CenterNet-master/models/ctdet_coco_dla_2x.pth 

运行以上面的demo代码时出现下面的错误（注意文件地址要换成你的地址）

Fix size testing.
training chunk_sizes: [32]
The output will be saved to  /home/vincent/Code/CenterNet-master/src/lib/../../exp/ctdet/default
heads {'hm': 80, 'wh': 2, 'reg': 2}
Creating model...
loaded /home/vincent/Code/CenterNet-master/models/ctdet_coco_dla_2x.pth, epoch 230
THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1535491974311/work/aten/src/THC/THCGeneral.cpp line=663 error=11 : invalid argument
Traceback (most recent call last):
  File "demo.py", line 56, in <module>
    demo(opt)
  File "demo.py", line 49, in demo
    ret = detector.run(image_name)
  File "/home/vincent/Code/CenterNet-master/src/lib/detectors/base_detector.py", line 116, in run
    output, dets, forward_time = self.process(images, return_time=True)
  File "/home/vincent/Code/CenterNet-master/src/lib/detectors/ctdet.py", line 30, in process
    output = self.model(images)[-1]
  File "/home/vincent/anaconda3/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/vincent/Code/CenterNet-master/src/lib/models/networks/pose_dla_dcn.py", line 481, in forward
    z[head] = self.__getattr__(head)(y[-1])
  File "/home/vincent/anaconda3/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/vincent/anaconda3/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/container.py", line 91, in forward
    input = module(input)
  File "/home/vincent/anaconda3/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/vincent/anaconda3/envs/CenterNet/lib/python3.6/site-packages/torch/nn/modules/conv.py", line 301, in forward
    self.padding, self.dilation, self.groups)
RuntimeError: CUDA error: out of memory

解决办法

在demo.py里加上如下内容

import torch
torch.backends.cudnn.enabled = False

参考：cuda out of memory #184