Segmentation fault (core dumped)

[fly520fly520]

My Code :

create example data,

x = torch.ones((1, 3, 112, 112)).cuda()

model_trt_new = TRTModule()

model_trt_new.load_state_dict(torch.load('face_trt.pth'))
y_trt_new = model_trt_new(x)

This code can run normally, but when it is finished, I will get an error message as follows：
Segmentation fault (core dumped)
And create a core.xxx file
I use gdb to open the core.xxx file and get the following information：
Core was generated by `python'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0 0x00007f9105d527d8 in ?? ()

[hmmlencat]

hello, i meet the same problem as you.
code can run normally, but when it is finished, I will get an error message as follows：
Segmentation fault
Also the results is not match with the torch's output
please give me some advice.

[comeeasy]

I have same issue too.. please solve it..

[Luizerko]

Hello!

I have the same problem as you guys. In my case it also happens when I import torch2trt library, regardless of what comes after. And actually the error varies a little bit, sometimes a Segmentation fault, but sometimes free(): invalid pointer or even munmap_chunk(): invalid pointer.

I hope someone can solve it.

PS: cuda 10.2, pythorch 1.1.0, tensorrt 7.0.0.11, torch2trt 0.3.0

[xinwangChen]

I have same issue too.