experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

fixed the problem of default rnn ops never use tensor cores

Browse Source
This commit is contained in:
Kaixi Hou 2019-05-28 17:23:59 -07:00
parent 53fd64291f
commit 660925846c
1 changed files with 15 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
tensorflow/stream_executor/cuda/cuda_dnn.cc
View File

@ -1087,15 +1087,23 @@ class CudnnRnnDescriptor : public dnn::RnnDescriptor {
// We can only reasonably expect the user to handle the subsequent failure
// in profile mode, which is run with algorithms returned from
// GetRnnAlgorithms() (which are non-default and explicitly set whether to
// use tensor ops).
if (RnnTensorOpMathEnabled() && algorithm_config.algorithm().has_value()) {
cudnnMathType_t math_type =
algorithm_config.algorithm()->tensor_ops_enabled()
? CUDNN_TENSOR_OP_MATH
: CUDNN_DEFAULT_MATH;
// use tensor ops). CuDNN 7.2.1 fixed this issue
if (RnnTensorOpMathEnabled()) {
cudnnMathType_t math_type;
if (algorithm_config.algorithm().has_value()) {
math_type = algorithm_config.algorithm()->tensor_ops_enabled()
? CUDNN_TENSOR_OP_MATH
: CUDNN_DEFAULT_MATH;
} else {
#if CUDNN_VERSION >= 7201
math_type = CUDNN_TENSOR_OP_MATH;
#else
math_type = CUDNN_DEFAULT_MATH;
#endif // CUDNN_VERSION >= 7201
}
CHECK_CUDNN_OK(cudnnSetRNNMatrixMathType(rnn_desc.get(), math_type));
}
#endif
#endif // CUDNN_VERSION >= 7000
return CudnnRnnDescriptor(cudnn, std::move(rnn_desc), std::move(rnn_plan),
num_layers, hidden_size, input_size, batch_size,