caffe 2 c++ distribution training on multi computer

[michael81420]

請問有知道相關的caffe2 c++ function 與 他官方給的python 類似用法嗎 [https://caffe2.ai/docs/distributed-training.html] 可以利用nfs 達到多台電腦訓練的效果

[breadbread1984]

就知道run_plan_mpi可以分布式训练

[fxj8911]

hi，Sir. run make_mnist_db --image_file=tmp/train-images-idx3-ubyte --label_file=tmp/train-labels-idx1-ubyte --output_file=mnist-train-nchw-leveldb --channel_first --db leveldb

and PC return Aborted at 1522396348 (unix time) try "date -d @1522396348" if you are using GNU date *** PC: @ 0x404714 caffe2::convert_dataset() *** SIGSEGV (@0x0) received by PID 31170 (TID 0x7fc6627af800) from PID 0; stack trace: *** @ 0x7fc660a6a4b0 (unknown) @ 0x404714 caffe2::convert_dataset() @ 0x40368e main @ 0x7fc660a55830 __libc_start_main @ 0x4036c9 _start @ 0x0 (unknown) Segmentation fault (core dumped) Do you know what's wrong with my caffe2 environment?

[breadbread1984]

this may caused by the absence of mnist dataset. please check out whether the mnist dataset has been download successfully.