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

[tf.data] Increase the roll out percentage of optimization map_parallelization to 100%.

Browse Source 
PiperOrigin-RevId: 343870179
Change-Id: I0fff3616e045149a883c3c9eb6538400ba2f7a4c
This commit is contained in:
Jay Shi 2020-11-23 09:50:07 -08:00 committed by TensorFlower Gardener
parent 1da597f01d
commit e2de99a4e3
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/data/optimize_dataset_op.cc
View File

@ -85,7 +85,7 @@ void OptimizeDatasetOp::MakeDataset(OpKernelContext* ctx, DatasetBase* input,
// clang-format off // clang-format off
absl::flat_hash_map<string, uint64> live_experiments = { absl::flat_hash_map<string, uint64> live_experiments = {
{"enable_gradient_descent", 0}, {"enable_gradient_descent", 0},
{"map_parallelization", 50} {"map_parallelization", 100}
}; };
// clang-format on // clang-format on
auto hash_func = [](const string& str) { return Hash64(str); }; auto hash_func = [](const string& str) { return Hash64(str); };