Skip to content

ocean1992/pytorch_docker

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 

Repository files navigation

pytorch_docker

My docker image file of pytorch

I use pytorch docker image from docker hub, tag is :0.4.1-cuda9-cudnn7-devel. Once I updated my code (a small update, just code refine, not related to pytorch, i think) and run. Strange Errors:

*** Error in `/opt/conda/bin/python': free(): invalid pointer: 0x0000556236118dd8 ***
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x777e5)[0x7f2d88e757e5]
/lib/x86_64-linux-gnu/libc.so.6(+0x8037a)[0x7f2d88e7e37a]
/lib/x86_64-linux-gnu/libc.so.6(cfree+0x4c)[0x7f2d88e8253c]

The errors have been discussed in github in pytorch issues including (#2314, #2339.

Someone provides a solution:

sudo apt-get install libtcmalloc-minimal4
export LD_PRELOAD="/usr/lib/libtcmalloc_minimal.so.4"

However when i add these two lines in my docker image file. New errors appear:

[libprotobuf FATAL ./google/protobuf/map_field.h:388] Protocol Buffer map usage error:
MapKey::type MapKey is not initialized. Call set methods to initialize MapKey.
terminate called after throwing an instance of 'google::protobuf::FatalException'
  what():  Protocol Buffer map usage error:
MapKey::type MapKey is not initialized. Call set methods to initialize MapKey.

Another one says install pytorch with pip can fix this bug (official pytorch docker image install using conda)

I write this docker image using pip to install torch and it works!

Just share and provide a way to fix this bug.

About

My docker image file of pytorch

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published