我正在尝试测试pix2pixHD。我没有GPU。我正在使用pytorch==0.4.1
。我正在使用以下命令运行城市模型测试:
python test.py --name label2city_1024p --netG local --ngf 32 --gpu_ids -1 --resize_or_crop none
但是我得到Illegal instruction (core dumped)
。我用gdb尝试了回溯,发现了这一点:
#0 0x00007fcd623175e6 in THFloatVector_normal_fill_AVX2 () from /home/momen/nvidia_vid2vid/lib/python3.6/site-packages/torch/lib/libcaffe2.so
#1 0x00007fcd621d06f0 in THFloatTensor_normal () from /home/momen/nvidia_vid2vid/lib/python3.6/site-packages/torch/lib/libcaffe2.so
#2 0x00007fcd61c85156 in at::CPUFloatType::normal_(at::Tensor&, double, double, at::Generator*) const ()
from /home/momen/nvidia_vid2vid/lib/python3.6/site-packages/torch/lib/libcaffe2.so
#3 0x00007fcdb8724895 in torch::autograd::VariableType::normal_ (this=0x27d4fc0, self=..., mean=0, std=0.02, generator=0x0)
at torch/csrc/autograd/generated/VariableType.cpp:10819
#4 0x00007fcdb89e6300 in at::Tensor::normal_ (generator=0x0, std=0.02, mean=0, this=0x7fcd921cc958) at /pytorch/torch/lib/tmp_install/include/ATen/TensorMethods.h:700
如何解决此问题?欢迎任何类型的帮助。谢谢。
其他信息:
我尝试了py-bt,而问题m.weight.data.normal_(0.0, 0.02)
出现在第一Conv2d
层。