Compilation fails with cuDNN and cuda 7.5.18

[vchuravy]

If I compile the current master with cuDNN enable I encounter many of these errors:

src/operator/./cudnn_rnn-inl.h(127): error: argument of type "mshadow::half::half_t *" is incompatible with parameter of type "cudnnTensorDescriptor_t"
          detected during:
            instantiation of "void mxnet::op::CuDNNRNNOp<DType>::Forward(const mxnet::OpContext &, const std::vector<mxnet::TBlob, std::allocator<mxnet::TBlob>> &, const std::vector<mxnet::OpReqType, std::allocator<mxnet::OpReqType>> &, const std::vector<mxnet::TBlob, std::allocator<mxnet::TBlob>> &, const std::vector<mxnet::TBlob, std::allocator<mxnet::TBlob>> &) [with DType=mshadow::half::half_t]" 
(54): here
            instantiation of "mxnet::op::CuDNNRNNOp<DType>::CuDNNRNNOp(mxnet::op::RNNParam) [with DType=mshadow::half::half_t]" 
src/operator/rnn.cu(20): here

Versions:

• gcc: 4.9.2
• cuDNN: 5.0.4 (according to header file)
• cuda: 7.5.18
• mxnet: b12f957

[mli]

try git submodule update or use cudnn 5.1

[shuokay]

This maybe caused by #2795, you should update cuda and cudnn to the latest version.

[vchuravy]

I have just encountered this again. Features that require specific cuda/cuDNN versions should check for them and either disable themselves or print a error message that prompts the user to update.

[phunterlau]

This issue is closed due to lack of activity in the last 90 days. Feel free to reopen if this is still an active issue. Thanks!