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

Fix uses of private, mangled names for proto enumerators.

Browse Source 
PiperOrigin-RevId: 256558252
This commit is contained in:
A. Unique TensorFlower 2019-07-04 08:57:16 -07:00 committed by TensorFlower Gardener
parent 1375ad20ad
commit 011dac5f0a
4 changed files with 6 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
tensorflow/compiler/jit/tests/auto_clustering_test_helper.cc
View File

@ -128,9 +128,7 @@ Status AutoClusteringTest::RunAutoClusteringTestImpl(
TF_RETURN_IF_ERROR(AssertGraphDefIsUnclustered(graphdef)); TF_RETURN_IF_ERROR(AssertGraphDefIsUnclustered(graphdef));
OptimizationPassRunner runner; OptimizationPassRunner runner;
TF_RETURN_IF_ERROR( TF_RETURN_IF_ERROR(runner.SetJitLevel(tensorflow::OptimizerOptions::ON_2));
runner.SetJitLevel(tensorflow::OptimizerOptions::GlobalJitLevel::
OptimizerOptions_GlobalJitLevel_ON_2));
TF_RETURN_IF_ERROR(runner.AddCpus(32)); TF_RETURN_IF_ERROR(runner.AddCpus(32));
TF_RETURN_IF_ERROR(runner.AddGpus(8)); TF_RETURN_IF_ERROR(runner.AddGpus(8));
@ -211,9 +209,7 @@ Status BenchmarkMarkForCompilation(absl::string_view graph_def_path,
ReadTextProto(Env::Default(), string(graph_def_path), &graph_def)); ReadTextProto(Env::Default(), string(graph_def_path), &graph_def));
OptimizationPassRunner runner; OptimizationPassRunner runner;
TF_RETURN_IF_ERROR( TF_RETURN_IF_ERROR(runner.SetJitLevel(tensorflow::OptimizerOptions::ON_2));
runner.SetJitLevel(tensorflow::OptimizerOptions::GlobalJitLevel::
OptimizerOptions_GlobalJitLevel_ON_2));
TF_RETURN_IF_ERROR(runner.AddCpus(32)); TF_RETURN_IF_ERROR(runner.AddCpus(32));
TF_RETURN_IF_ERROR(runner.AddGpus(8)); TF_RETURN_IF_ERROR(runner.AddGpus(8));

3
tensorflow/compiler/tf2xla/sharding_util_test.cc
View File

@ -25,8 +25,7 @@ TEST(CoreUtilTest, ParseShardingFromDevice) {
auto core_from_sharding = auto core_from_sharding =
[](absl::optional<xla::OpSharding> sharding) -> int64 { [](absl::optional<xla::OpSharding> sharding) -> int64 {
if (sharding.has_value() && if (sharding.has_value() &&
sharding.value().type() == sharding.value().type() == xla::OpSharding::MAXIMAL) {
xla::OpSharding::Type::OpSharding_Type_MAXIMAL) {
return sharding.value().tile_assignment_devices(0); return sharding.value().tile_assignment_devices(0);
} else { } else {
return -1; return -1;

3
tensorflow/compiler/tf2xla/tf2xla_util.cc
View File

@ -504,8 +504,7 @@ Status SetNodeShardingFromNeighbors(Node* n, bool out_edges) {
*possible_match, *possible_match,
/*num_cores_per_replica=*/std::numeric_limits<int32>::max())); /*num_cores_per_replica=*/std::numeric_limits<int32>::max()));
if (sharding.has_value()) { if (sharding.has_value()) {
TF_RET_CHECK(sharding.value().type() == TF_RET_CHECK(sharding.value().type() == xla::OpSharding::MAXIMAL);
xla::OpSharding::Type::OpSharding_Type_MAXIMAL);
const int core_annotation = sharding.value().tile_assignment_devices(0); const int core_annotation = sharding.value().tile_assignment_devices(0);
if (core == -1 || core > core_annotation) { if (core == -1 || core > core_annotation) {
core = core_annotation; core = core_annotation;

5
tensorflow/compiler/tf2xla/xla_compiler.cc
View File

@ -85,8 +85,7 @@ ComputeArgAndRetvalCores(const Graph& graph) {
auto sharding, auto sharding,
ParseShardingFromDevice(*n, std::numeric_limits<int32>::max())); ParseShardingFromDevice(*n, std::numeric_limits<int32>::max()));
if (sharding.has_value()) { if (sharding.has_value()) {
TF_RET_CHECK(sharding.value().type() == TF_RET_CHECK(sharding.value().type() == xla::OpSharding::MAXIMAL);
xla::OpSharding::Type::OpSharding_Type_MAXIMAL);
return sharding.value().tile_assignment_devices(0); return sharding.value().tile_assignment_devices(0);
} else { } else {
return -1; return -1;
@ -832,7 +831,7 @@ Status XlaCompiler::BuildArguments(
xla::XlaOp tuple; xla::XlaOp tuple;
if (is_entry_computation) { if (is_entry_computation) {
xla::OpSharding tuple_sharding; xla::OpSharding tuple_sharding;
tuple_sharding.set_type(xla::OpSharding::Type::OpSharding_Type_TUPLE); tuple_sharding.set_type(xla::OpSharding::TUPLE);
for (int64 parameter : *input_to_args) { for (int64 parameter : *input_to_args) {
auto it = arg_cores.find(parameter); auto it = arg_cores.find(parameter);
const int core = it == arg_cores.end() ? 0 : it->second; const int core = it == arg_cores.end() ? 0 : it->second;