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

Make Softplus benchmark more meaningful by using random inputs. The speed is highly input value dependent.

Browse Source 
PiperOrigin-RevId: 263034339
This commit is contained in:
A. Unique TensorFlower 2019-08-12 16:54:09 -07:00 committed by TensorFlower Gardener
parent 090ad6034d
commit 41e3677707
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/nn_ops_test.cc
View File

@ -1277,7 +1277,7 @@ static void BM_SoftplusFloat(int iters, int batch_size, int rows, int cols,
gtl::InlinedVector<TensorValue, 4> inputs; gtl::InlinedVector<TensorValue, 4> inputs;
TensorShape shape1({batch_size, rows, cols, depth}); TensorShape shape1({batch_size, rows, cols, depth});
Tensor input1(DT_FLOAT, shape1); Tensor input1(DT_FLOAT, shape1);
test::FillIota<float>(&input1, 1.0); input1.flat<float>().setRandom();
inputs.push_back({nullptr, &input1}); inputs.push_back({nullptr, &input1});
// Softplusing op. // Softplusing op.