caffe_copy ERROR cudaSuccess (11 vs 0)

[hyojinie]

Hi I don't know why caffe_copy gives me error == cudaSuccess (11 vs 0) invalid argument.

*** Check failure stack trace: ***
@ 0x7f81c6149c3c google::LogMessage::Fail()
@ 0x7f81c6149b88 google::LogMessage::SendToLog()
@ 0x7f81c614958a google::LogMessage::Flush()
@ 0x7f81c614c521 google::LogMessageFatal::~LogMessageFatal()
@ 0x7f81c675a10b caffe::caffe_copy<>()
@ 0x7f81c67e6f95 caffe::NetvladLayer<>::Forward_cpu()
@ 0x7f81c687c561 caffe::Net<>::ForwardFromTo()
@ 0x7f81c687c8d7 caffe::Net<>::ForwardPrefilled()
@ 0x7f81c68b48e9 caffe::Solver<>::Test()
@ 0x7f81c68b524e caffe::Solver<>::TestAll()
@ 0x7f81c68b5395 caffe::Solver<>::Step()
@ 0x7f81c68b5ea5 caffe::Solver<>::Solve()
@ 0x409ae3 train()
@ 0x407101 main
@ 0x7f81c5665ec5 (unknown)
@ 0x407811 (unknown)
Aborted (core dumped)

The error occurs here.

caffe_copy(dim, bottom_data + n * dim, residual_data);

dim, n: int
bottom_data, residual_data: Dtype pointer

I also printed the arguments.
I0227 22:35:27.273406 4808 vlad_layer.cpp:175] dim 43264
I0227 22:35:27.273412 4808 vlad_layer.cpp:176] n 0
I0227 22:35:27.273418 4808 vlad_layer.cpp:177] bottom_data 0x21bfc0000
I0227 22:35:27.273428 4808 vlad_layer.cpp:178] residual_data 0x209935400

I use titian X for my GPU. and CUDA 7.5, and the Caffe downloaded in January 2016.

[seanbell]

NetvladLayer? Is this an error in a layer you wrote yourself? If so, please ask for help on the mailing list, and close this issue. Github issues are not for discussing modeling questions, or for helping with debugging code you wrote yourself.

Please do not post usage, installation, or modeling questions, or other requests for help to Issues.
Use the caffe-users list instead. This helps developers maintain a clear, uncluttered, and efficient view of the state of Caffe.

[hyojinie]

I think my problem occurred because I my forward_gpu just calls forward_cpu, and so the caffe_copy gets cpu pointers while running in the gpu mode.

[seanbell]

Aah yes that's a common mistake to make; glad you fixed it!