centos7.3 gpu Not supported

[foxgaga]

CentOS Linux release 7.3.1611 (Core)
cuda-8.0
cudnn v5

https://github.com/PaddlePaddle/book/blob/develop/05.recommender_system/train.py

alter

paddle.init(use_gpu=True, gpu_id=1)

output

I0712 21:03:33.468981 1598729 Util.cpp:166] commandline:  --use_gpu=True --gpu_id=1
I0712 21:03:36.654531 1598729 GradientMachine.cpp:85] Initing parameters..
I0712 21:03:36.688349 1598729 GradientMachine.cpp:92] Init parameters done.
F0712 21:03:36.766964 1598729 Matrix.cpp:645] Not supported
*** Check failure stack trace: ***
    @     0x7efe05c6406d  google::LogMessage::Fail()
    @     0x7efe05c65eef  google::LogMessage::SendToLog()
    @     0x7efe05c63be3  google::LogMessage::Flush()
    @     0x7efe05c6680e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7efe05abf586  paddle::GpuMatrix::mul()
    @     0x7efe059856b5  paddle::FullyConnectedLayer::forward()
    @     0x7efe05a02c20  paddle::NeuralNetwork::forward()
    @     0x7efe059f4eb3  paddle::GradientMachine::forwardBackward()
    @     0x7efe05c3fbcb  GradientMachine::forwardBackward()
    @     0x7efe058b74ce  _wrap_GradientMachine_forwardBackward
    @     0x7efe0f442412  PyEval_EvalFrameEx
    @     0x7efe0f4441ce  PyEval_EvalCodeEx
    @     0x7efe0f4431f6  PyEval_EvalFrameEx
    @     0x7efe0f4441ce  PyEval_EvalCodeEx
    @     0x7efe0f4431f6  PyEval_EvalFrameEx
    @     0x7efe0f4441ce  PyEval_EvalCodeEx
    @     0x7efe0f4431f6  PyEval_EvalFrameEx
    @     0x7efe0f443323  PyEval_EvalFrameEx
    @     0x7efe0f4441ce  PyEval_EvalCodeEx
    @     0x7efe0f4442e2  PyEval_EvalCode
    @     0x7efe0f464960  PyRun_FileExFlags
    @     0x7efe0f464b3f  PyRun_SimpleFileExFlags
    @     0x7efe0f47a484  Py_Main
    @     0x7efe0e681b35  __libc_start_main
    @           0x400649  (unknown)
Aborted

[llxxxll]

Single GPU settings
paddle.init(use_gpu=True, gpu_id=0)

[qingqing01]

For the sparse input data, the data can not be passed from Python to C++ when using one GPU. The code is here: https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/py_paddle/dataprovider_converter.py#L178 . I will try to fix it later. Thanks for your question.

[mical892524]

I have the same issue, how to fix this.

[Superjomn]

Too long ago, free to reopen it.