)

# Add Keras module aliases
_losses = _KerasLazyLoader(globals(), submodule="losses", name="losses")
_metrics = _KerasLazyLoader(globals(), submodule="metrics", name="metrics")
_optimizers = _KerasLazyLoader(
    globals(), submodule="optimizers", name="optimizers")
_initializers = _KerasLazyLoader(
    globals(), submodule="initializers", name="initializers")
setattr(_current_module, "losses", _losses)

    tensorflow::MarkForCompilationPassFlags* flags =
        tensorflow::GetMarkForCompilationPassFlags();
    flags->tf_xla_cpu_global_jit = true;
    flags->tf_xla_min_cluster_size = 1;
  } else {
    optimizer_options->set_global_jit_level(tensorflow::OptimizerOptions::OFF);
  }
}

unsigned char TF_SetXlaEnableLazyCompilation(unsigned char enable) {
  tensorflow::BuildXlaOpsPassFlags* flags =

        "//tensorflow/core:test_main",
    ],
)

tf_cc_test(
    name = "parallel_device_remote_test",
    srcs = ["parallel_device_remote_test.cc"],
    # TODO(b/136478427): Enable global heap checking when servers shut down
    # cleanly.
    args = ["--heap_check="],
    deps = [
        ":parallel_device",
        ":parallel_device_testlib",
        "//tensorflow/c:c_api",

  TF_Buffer* buffer = TF_NewBuffer();

  TF_GraphGetOpDef(graph, "Add", buffer, status);
  ASSERT_EQ(TF_OK, TF_GetCode(status));
  const OpDef* expected_op_def;
  TF_ASSERT_OK(OpRegistry::Global()->LookUpOpDef("Add", &expected_op_def));
  string expected_serialized;
  expected_op_def->SerializeToString(&expected_serialized);
  string actual_string(reinterpret_cast<const char*>(buffer->data),

          NodeDef* const node_def = graph_def.add_node();
          *node_def = node->def();
        }
      }
      *graph_def.mutable_library() = graph.flib_def().ToProto();
      if (flags::Global().more_stack_traces.value()) {
        *graph_def.mutable_debug_info() = graph.BuildDebugInfo();
      }
      session->graph->mu.unlock();
      status->status = session->session->Extend(std::move(graph_def));

      * Replace `tf.keras.mixed_precision.experimental.set_policy` with `tf.keras.mixed_precision.set_global_policy`. The experimental symbol `set_policy` was renamed to `set_global_policy` in the non-experimental API.

# configure script unconditionally); otherwise if it is false or undefined, the
# build is static and TensorFlow symbols (in Python only) are loaded into the
# global symbol table in order to support op registration. This means that
# projects building with Bazel and importing TensorFlow as a dependency will not
# depend on libtensorflow_framework.so unless they opt in.
tf_cc_shared_library(

    gcc_version = run_shell([gcc_env, '--version']).split()
    if gcc_version[0] in ('gcc', 'g++'):
      return gcc_env
  return None


def main():
  global _TF_WORKSPACE_ROOT
  global _TF_BAZELRC
  global _TF_CURRENT_BAZEL_VERSION

  parser = argparse.ArgumentParser()
  parser.add_argument(
      '--workspace',
      type=str,

    case DT_RESOURCE:
    case DT_VARIANT:
      return true;
    case DT_QINT8:
    case DT_QINT16:
    case DT_QINT32:
    case DT_QUINT8:
    case DT_QUINT16:
      return tensorflow::flags::Global()
          .enable_quantized_dtypes_training.value();
    default:
      return false;
  }
}

template <typename Gradient, typename BackwardFunction, typename TapeTensor>

build:ios_fat --config=ios
build:ios_fat --ios_multi_cpus=armv7,arm64,i386,x86_64

# Config to use a mostly-static build and disable modular op registration
# support (this will revert to loading TensorFlow with RTLD_GLOBAL in Python).
# By default, TensorFlow will build with a dependence on
# //tensorflow:libtensorflow_framework.so.
build:monolithic --define framework_shared_object=false
build:monolithic --define tsl_protobuf_header_only=false