Caffe训练模型时core dump

在安装好的 caffe 环境里训练模型时报错

$ cd <caffe根目录>
$ ./build/tools/caffe train -solver examples/mnist/lenet_solver.prototxt

...
I0111 15:27:57.013530 20030 layer_factory.hpp:77] Creating layer mnist
F0111 15:27:57.013805 20030 db_lmdb.hpp:15] Check failed: mdb_status == 0 (2 vs. 0) No such file or directory
*** Check failure stack trace: ***
    @     0x7f4ca2072e6d  (unknown)
    @     0x7f4ca2074ced  (unknown)
    @     0x7f4ca2072a5c  (unknown)
    @     0x7f4ca207563e  (unknown)
    @     0x7f4ca2554321  caffe::db::LMDB::Open()
    @     0x7f4ca24639b4  caffe::DataLayer<>::DataLayer()
    @     0x7f4ca2463b62  caffe::Creator_DataLayer<>()
    @     0x7f4ca252a149  caffe::Net<>::Init()
    @     0x7f4ca252c322  caffe::Net<>::Net()
    @     0x7f4ca2533a30  caffe::Solver<>::InitTrainNet()
    @     0x7f4ca2534983  caffe::Solver<>::Init()
    @     0x7f4ca2534c5f  caffe::Solver<>::Solver()
    @     0x7f4ca254bad1  caffe::Creator_SGDSolver<>()
    @           0x40ee8e  caffe::SolverRegistry<>::CreateSolver()
    @           0x40812c  train()
    @           0x405b8c  main
    @     0x7f4c9c8ddb35  __libc_start_main
    @           0x40654b  (unknown)
Aborted (core dumped)

问题原因是没有找到训练用的数据，所以我们需要先下载训练数据，如下

$ cd <caffe根目录>
$ ./data/mnist/get_mnist.sh
$ ./examples/mnist/create_mnist.sh

然后重新运行训练模型

$ cd <caffe根目录>

$ ./build/tools/caffe train -solver examples/mnist/lenet_solver.prototxt

或

$ ./examples/mnist/train_lenet.sh