facebookarchive / caffe2

Caffe2 is a lightweight, modular, and scalable deep learning framework.
https://caffe2.ai
Apache License 2.0
8.42k stars 1.95k forks source link

Build fails with Nvidia Jetson TK1 #423

Open rachitgarg91 opened 7 years ago

rachitgarg91 commented 7 years ago

Hello,

Build seems to be failing with nvidia Jetson TK1. Here is the log:

[ 40%] Building CXX object caffe2/CMakeFiles/Caffe2_CPU.dir/operators/utility_ops.cc.o [ 40%] Building CXX object caffe2/CMakeFiles/Caffe2_CPU.dir/operators/local_response_normalization_op.cc.o [ 41%] Building CXX object caffe2/CMakeFiles/Caffe2_CPU.dir/operators/rank_loss_op.cc.o [ 41%] Building CXX object caffe2/CMakeFiles/Caffe2_CPU.dir/operators/fully_connected_op.cc.o c++: internal compiler error: Killed (program cc1plus) Please submit a full bug report, with preprocessed source if appropriate. See file:///usr/share/doc/gcc-4.8/README.Bugs for instructions. make[2]: [caffe2/CMakeFiles/Caffe2_CPU.dir/operators/utility_ops.cc.o] Error 4 make[2]: Waiting for unfinished jobs.... make[1]: [caffe2/CMakeFiles/Caffe2_CPU.dir/all] Error 2 make: [all] Error 2

Thanks.

bwasti commented 7 years ago

This seems to be a compiler error. Perhaps you are running out of memory during the build? Please post the contents of file:///usr/share/doc/gcc-4.8/README.Bugs

xhzzhang commented 7 years ago

Hello rachitgarg91, the NVRTC library and header file is missed in my TX1, can you provide this file to me(xhzzhang@yeah.net), thanks!

rachitgarg91 commented 7 years ago

@bwasti It seems the file file:///usr/share/doc/gcc-4.8/README.Bugs is just a generic instruction on how to file gcc bugs. https://www.dropbox.com/s/ez08ap4eeoxojng/README.Bugs?dl=0

@xhzzhang I am running caffe2 on nvidia Jetson TK1, will the NVRTC library/header file be same for TX1? Let me know if you need it, I can share it with you.

xhzzhang commented 7 years ago

@rachitgarg91 In my opinion, NVRTC just match cuda toolkit, i use cuda-7.0 for ubuntu 14.04, if you have same cuda and ubuntu version, please share it, thanks.