some problems when training on my own dataset

[niezhongliang]

[niezhongliang]

Hi, is there any one met this situation: terminate called after throwing an instance of 'std::bad_alloc' what(): std::bad_alloc *** Aborted at 1544160602 (unix time) try "date -d @1544160602" if you are using GNU date *** PC: @ 0x7f5fb6607428 gsignal *** SIGABRT (@0x3e8000049ff) received by PID 18943 (TID 0x7f5f837fe700) from PID 18943; stack trace: *** @ 0x7f5fb66074b0 (unknown) @ 0x7f5fb6607428 gsignal @ 0x7f5fb660902a abort @ 0x7f5fb741b84d __gnu_cxx::__verbose_terminate_handler() @ 0x7f5fb74196b6 (unknown) @ 0x7f5fb7419701 std::terminate() @ 0x7f5fb7419919 __cxa_throw @ 0x7f5fb7442082 std::__throw_bad_alloc() @ 0x7f5fb860ecf4 std::vector<>::vector() @ 0x7f5fb86534d7 caffe::DetectionDataLayer<>::load_batch() @ 0x7f5fb86e5112 caffe::BasePrefetchingDataLayer<>::InternalThreadEntry() @ 0x7f5fb850b6e5 caffe::InternalThread::entry() @ 0x7f5fb850c60e boost::detail::thread_data<>::run() @ 0x7f5fac1eb5d5 (unknown) @ 0x7f5fa682a6ba start_thread @ 0x7f5fb66d941d clone @ 0x0 (unknown) Aborted (core dumped)

I use my data follow window_file format. And set the batch_size = 1,