Error: out of memory

[TonyDandelion]

Hello, Can you tell me what is your experimental facility I used a GTX1060 pc and a GTX1080TI pc for training network. Both of these computers can not carry on the experiment. Show the same errors as follow:
I0318 16:34:30.635128 11119 solver.cpp:351] Iteration 0, Testing net (#0)
F0318 16:34:30.773664 11119 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
@ 0x7fd1d8e915cd google::LogMessage::Fail()
@ 0x7fd1d8e93433 google::LogMessage::SendToLog()
@ 0x7fd1d8e9115b google::LogMessage::Flush()
@ 0x7fd1d8e93e1e google::LogMessageFatal::~LogMessageFatal()
@ 0x7fd1d9615488 caffe::SyncedMemory::mutable_gpu_data()
@ 0x7fd1d961fd12 caffe::Blob<>::mutable_gpu_data()
@ 0x7fd1d968c6f3 caffe::EltwiseLayer<>::Forward_gpu()
@ 0x7fd1d96626b1 caffe::Net<>::ForwardFromTo()
@ 0x7fd1d96627b7 caffe::Net<>::Forward()
@ 0x7fd1d960d292 caffe::Solver<>::Test()
@ 0x7fd1d960db6e caffe::Solver<>::TestAll()
@ 0x7fd1d96110c7 caffe::Solver<>::Step()
@ 0x7fd1d961138a caffe::Solver<>::Solve()
@ 0x40a914 train()
@ 0x407548 main
@ 0x7fd1d7627830 __libc_start_main
@ 0x407e19 _start
@ (nil) (unknown)
Aborted (core dumped)

And I had ready modify the bath_size from 128 to 12.

PS:Because of the limit computer memory(only 16GB). It seems that it's impossible to process all training pictures in floder which named as "Train_291". I used only 592(74*8) pictures to generate the train dataset that named as x234.h5. In fact, there are 2328(291*8) pictures in training dataset floder. I have no ideal whether it would influence the result in end if or not.