Segmentation fault
wgh1989 opened this issue · comments
env: torch-v1.2.0, cuda-10.1,
I got "Segmentation fault (core dumped)" when I try to use "scripts/pose/train_g8.sh" to train on the example pose data
when change to "scripts/face/train_g8_512.sh" is the same.
I have located that the error occured when I use "--warp_ref --spade_combine " options.
can somebody help me? what can I do to solve this problem?
ok, I have solved.
It is related to the base env.
The new env I used is "torch-1.2.0, cuda-10.0, apex with cuda_ext, gcc-540"