caffe训练时报错

转自：https://www.cnblogs.com/haiyang21/p/7614669.html

F0717 15:17:20.254209  8400 math_functions.cu:121] Check failed: status == CUBLAS_STATUS_SUCCESS (11 vs. 0)  CUBLAS_STATUS_MAPPING_ERROR
*** Check failure stack trace: ***
    @     0x7fbe7ed3a5cd  google::LogMessage::Fail()
    @     0x7fbe7ed3c433  google::LogMessage::SendToLog()
    @     0x7fbe7ed3a15b  google::LogMessage::Flush()
    @     0x7fbe7ed3ce1e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7fbe7f5563ca  caffe::caffe_gpu_asum<>()
    @     0x7fbe7f54a8a7  caffe::SoftmaxWithLossLayer<>::Forward_gpu()
    @     0x7fbe7f4d2a82  caffe::Net<>::ForwardFromTo()
    @     0x7fbe7f4d2ba7  caffe::Net<>::Forward()
    @     0x7fbe7f4d2dac  caffe::Net<>::Forward()
    @     0x7fbe7f4b9dfd  caffe::Solver<>::Step()
    @     0x7fbe7f4ba969  caffe::Solver<>::Solve()
    @           0x40c7a7  train()
    @           0x4085d8  main
    @     0x7fbe7db0e830  __libc_start_main
    @           0x408d49  _start
    @              (nil)  (unknown)

原因：训练的标签list不对，如 用18类的标签list训练2类的模型