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

Also set the unroll factor in the C++ integration code for tanh.

Browse Source 
PiperOrigin-RevId: 317083225
Change-Id: I7c2c26d664c15cbc967188da4b3012161edbcf49
This commit is contained in:
Adrian Kuegel 2020-06-18 05:26:02 -07:00 committed by TensorFlower Gardener
parent 7378fabf90
commit 78f6d51b8d
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorflow/core/kernels/mlir_generated_cwise_op_gpu_tanh.cu.cc
View File

@ -124,7 +124,7 @@ class MlirGeneratedTanhOp : public OpKernel {
// This has to be aligned with the configuration that was used when building
// the kernels. See the corresponding build rules in `cubin_headers/BUILD`.
LaunchConfig config = GetLaunchConfiguration(
{256}, {}, {static_cast<uint64>(inp.NumElements())});
{256}, {4}, {static_cast<uint64>(inp.NumElements())});
OP_REQUIRES_OK(
ctx, stream->parent()->Launch(stream, config.threadDim, config.blockDim,
*kernel, args));