I am having crashes at import time when trying to run pytorch through pytest unit tests in IDE, with no meaningful information in the stack trace:
❯ coredumpctl info 12120 *
PID: 12120 (python)
UID: 10022 (XXXXXXXXX)
GID: 1000 (XXXXXXXXX)
Signal: 6 (ABRT)
Timestamp: Thu 2018-09-13 14:09:05 CEST (42min ago)
Command Line: /home/XXXXXXXXX/.env/bin/python /usr/share/java/pycharm-community/helpers/pycharm/_jb_pytest_runner.py --target test_pytorch.py::test_pool_01
Executable: /usr/bin/python3.6
Control Group: /user.slice/user-10022.slice/session-3.scope
Unit: session-3.scope
Slice: user-10022.slice
Session: 3
Owner UID: 10022 (XXXXXXXXX)
Boot ID: c5c915cb3ad04176800e940692d66625
Machine ID: 9ad705d336bc443abb3ac948d7d2987c
Hostname: XXXXXX
Storage: /var/lib/systemd/coredump/core.python.10022.c5c915cb3ad04176800e940692d66625.12120.1536840545000000.lz4
Message: Process 12120 (python) of user 10022 dumped core.
Stack trace of thread 12120:
#0 0x00007fcff1be1e5b raise (libpthread.so.0)
#1 0x00007fcff1be1fc0 __restore_rt (libpthread.so.0)
#2 0x00007fcff10acfeb raise (libc.so.6)
#3 0x00007fcff10975c1 abort (libc.so.6)
#4 0x00007fcf8ba19a9b _ZN9__gnu_cxx27__verbose_terminate_handlerEv.cold.1 (libstdc++.so.6)
#5 0x00007fcf8ba1fefc _ZN10__cxxabiv111__terminateEPFvvE (libstdc++.so.6)
#6 0x00007fcf8ba1ff57 _ZSt9terminatev (libstdc++.so.6)
#7 0x00007fcf8ba201b8 __cxa_throw (libstdc++.so.6)
#8 0x00007fcfb7f2ca61 n/a (/home/XXXXXXXXX/.env/lib/python3.6/site-packages/torch/_C.cpython-36m-x86_64-linux-gnu.so)
Does the pytorch project or someone else publish debug wheels somewhere ? I can see many build jobs at https://ci.pytorch.org but no published artifacts.