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

Separate code path for cuDNN 8 to set RNN descriptor.

Browse Source 
The old API has been deprecated, and cudnnSetRNNMatrixMathType returns an error.

PiperOrigin-RevId: 319277796
Change-Id: I793769abf537fbfe13b44b381318184e0d8b89f3
This commit is contained in:
Christian Sigg 2020-07-01 12:41:44 -07:00 committed by TensorFlower Gardener
parent 9214e6dac3
commit 28766652e6
1 changed files with 43 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

72
tensorflow/stream_executor/cuda/cuda_dnn.cc
View File

@ -1091,12 +1091,55 @@ class CudnnRnnDescriptor : public dnn::RnnDescriptor {
if (use_projection) { if (use_projection) {
unified_size = cell_size; unified_size = cell_size;
} }
// Require explicit algorithm config to enable tensor cores. Some configs
// return CUDNN_NOT_SUPPORTED when tensor ops are enabled (which is against
// the idiom that enabling tensor ops is only a hint: see nvbugs/2172799).
// 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). CuDNN 7.2.1 fixed this issue.
// TODO(csigg): Minimal support cuDNN version is 7.3, clean up.
bool allow_tensor_ops =
data_type != CUDNN_DATA_FLOAT || tensorflow::tf32_execution_allowed();
bool use_tensor_ops =
algorithm_config.algorithm().has_value()
? algorithm_config.algorithm()->tensor_ops_enabled()
: allow_tensor_ops;
if (use_tensor_ops && !allow_tensor_ops) {
return port::Status(port::error::INVALID_ARGUMENT,
"Algo requests disallowed tensor op evaluation.");
}
#if CUDNN_VERSION >= 8000
cudnnMathType_t math_type =
use_tensor_ops ? CUDNN_TENSOR_OP_MATH : CUDNN_FMA_MATH;
cudnnRNNBiasMode_t bias_mode = CUDNN_RNN_DOUBLE_BIAS;
uint32_t aux_flags = 0;
if (use_padded_io) aux_flags |= CUDNN_RNN_PADDED_IO_ENABLED;
RETURN_IF_CUDNN_ERROR(cudnnSetRNNDescriptor_v8(
/*rnnDesc=*/rnn_desc.get(), /*algo=*/rnn_algo, /*cellMode=*/rnn_mode,
/*biasMode=*/bias_mode, /*dirMode=*/direction_mode,
/*inputMode=*/input_mode,
/*dataType=*/data_type, /*mathPrec=*/compute_type,
/*mathType=*/math_type,
/*inputSize=*/input_size,
/*hiddenSize=*/hidden_size, /*projSize=*/cell_size,
/*numLayers=*/num_layers,
/*dropoutDesc=*/dropout_desc.handle(),
/*auxFlags=*/aux_flags));
#else
cudnnMathType_t math_type =
use_tensor_ops ? CUDNN_TENSOR_OP_MATH : CUDNN_DEFAULT_MATH;
RETURN_IF_CUDNN_ERROR(cudnnSetRNNDescriptor_v6( RETURN_IF_CUDNN_ERROR(cudnnSetRNNDescriptor_v6(
cudnn.handle(), /*rnnDesc=*/rnn_desc.get(), cudnn.handle(), /*rnnDesc=*/rnn_desc.get(),
/*hiddenSize=*/unified_size, /*numLayers=*/num_layers, /*hiddenSize=*/unified_size, /*numLayers=*/num_layers,
/*dropoutDesc=*/dropout_desc.handle(), /*inputMode=*/input_mode, /*dropoutDesc=*/dropout_desc.handle(), /*inputMode=*/input_mode,
/*direction=*/direction_mode, /*mode=*/rnn_mode, /*algo=*/rnn_algo, /*direction=*/direction_mode, /*mode=*/rnn_mode, /*algo=*/rnn_algo,
/*dataType=*/compute_type)); /*dataType=*/compute_type));
CHECK_CUDNN_OK(cudnnSetRNNMatrixMathType(rnn_desc.get(), math_type));
#endif
if (use_projection) { if (use_projection) {
RETURN_IF_CUDNN_ERROR(cudnnSetRNNProjectionLayers( RETURN_IF_CUDNN_ERROR(cudnnSetRNNProjectionLayers(
cudnn.handle(), /*rnnDesc=*/rnn_desc.get(), cudnn.handle(), /*rnnDesc=*/rnn_desc.get(),
@ -1132,35 +1175,6 @@ class CudnnRnnDescriptor : public dnn::RnnDescriptor {
cudnn, input_size, data_type, rnn_desc.get(), cudnn, input_size, data_type, rnn_desc.get(),
rnn_mode, direction_mode, num_layers)); rnn_mode, direction_mode, num_layers));
// Require explicit algorithm config to enable tensor cores. Some configs
// return CUDNN_NOT_SUPPORTED when tensor ops are enabled (which is against
// the idiom that enabling tensor ops is only a hint: see nvbugs/2172799).
// 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). CuDNN 7.2.1 fixed this issue
bool allow_tensor_ops =
data_type != CUDNN_DATA_FLOAT || tensorflow::tf32_execution_allowed();
bool use_tensor_ops;
if (algorithm_config.algorithm().has_value()) {
use_tensor_ops = algorithm_config.algorithm()->tensor_ops_enabled();
} else {
use_tensor_ops = allow_tensor_ops;
}
if (use_tensor_ops && !allow_tensor_ops) {
return port::Status(port::error::INVALID_ARGUMENT,
"Algo requests disallowed tensor op evaluation.");
}
cudnnMathType_t math_type;
#if CUDNN_VERSION >= 8000
math_type = use_tensor_ops ? CUDNN_TENSOR_OP_MATH : CUDNN_FMA_MATH;
#else
math_type = use_tensor_ops ? CUDNN_TENSOR_OP_MATH : CUDNN_DEFAULT_MATH;
#endif
CHECK_CUDNN_OK(cudnnSetRNNMatrixMathType(rnn_desc.get(), math_type));
return CudnnRnnDescriptor(cudnn, std::move(rnn_desc), std::move(rnn_plan), return CudnnRnnDescriptor(cudnn, std::move(rnn_desc), std::move(rnn_plan),
num_layers, hidden_size, input_size, cell_size, num_layers, hidden_size, input_size, cell_size,
batch_size, input_mode, direction_mode, rnn_mode, batch_size, input_mode, direction_mode, rnn_mode,