//
// Below here we do the gradient algorithm. It works as follows:
//
// First we filter the tape to just the subset of operations we want to
// differentiate. In the process of doing so we count how many times each Tensor
// is used as an input to an op (so we know when we're done computing gradients
// for that Tensor). We also count, for each tape entry, how many of its output

test:rbe_base --test_env=USER=anon

# TODO(kanglan): Check if we want to merge rbe_linux into rbe_linux_cpu.
build:rbe_linux --config=rbe_base
build:rbe_linux --action_env=PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/local/go/bin"
# Non-rbe settings we should include because we do not run configure
build:rbe_linux --config=avx_linux
# TODO(gunan): Check why we need this specified in rbe, but not in other builds.

  gpu_options->set_allow_growth(gpu_memory_allow_growth);

  (*config.mutable_device_count())["CPU"] = num_cpu_devices;

  // TODO(b/113217601): This is needed for EagerContext::runner_ to use a
  // threadpool, so that we avoid the possibility of running the runner_ in the
  // threadpool of GPU event mgr, as that can trigger more callbacks to be
  // scheduled on that same threadpool, causing a deadlock in cases where the

# buildifier: disable=load-on-top

workspace(name = "org_tensorflow")

# buildifier: disable=load-on-top

# We must initialize hermetic python first.
load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")

http_archive(
    name = "bazel_skylib",
    sha256 = "74d544d96f4a5bb630d465ca8bbcfe231e3594e5aae57e1edbf17a6eb3ca2506",
    urls = [

          # returned by the API, but a sha256sum of the entire chunk of image
          # metadata. gcr.io helpfully includes it in the header of the response
          # as docker-content-digest: sha256:[digest]. Note we use egrep to
          # match exactly sha256:<hash> because curl may include a ^M symbol at
          # the end of the line.

def get_ndk_api_level(environ_cp, android_ndk_home_path):
  """Gets the appropriate NDK API level to use for the provided Android NDK path.
  """

  # First check to see if we're using a blessed version of the NDK.
  properties_path = '%s/source.properties' % android_ndk_home_path
  if is_windows() or is_cygwin():
    properties_path = cygpath(properties_path)
  with open(properties_path, 'r') as f:

*   Added Text Dashboard to TensorBoard.

## Deprecations

*   TensorFlow 1.1.0 will be the last time we release a binary with Mac GPU
    support. Going forward, we will stop testing on Mac GPU systems. We continue
    to welcome patches that maintain Mac GPU support, and we will try to keep
    the Mac GPU build working.

## Changes to contrib APIs

        tensorflow::unwrap(handles[i]);
    if (tensorflow::CustomDeviceTensorHandle::classof(unwrapped_handle)) {
      // One of the inputs we're trying to pack is on a custom device. We'll let
      // the first custom device we see handle all of the packing.
      auto* custom_device_handle =
          tensorflow::down_cast<tensorflow::CustomDeviceTensorHandle*>(
              unwrapped_handle);

)

config_setting(
    name = "force_rtti",
    define_values = {
        "tf_force_rtti": "true",
    },
    visibility = ["//visibility:public"],
)

# Config setting for determining if we are building for Android.
config_setting(
    name = "android",
    constraint_values = if_google(
        ["//third_party/bazel_platforms/os:android"],
        [],
    ),
    values = if_oss(

    licenses = ["notice"],
)

# Currently pybind extension shared objects must use only C API headers since
# the C API has static initializers duplicated in the Python bindings. So we
# need a second rule that omits .cc files, in
# tensorflow/python:_pywrap_parallel_device.
filegroup(
    name = "lib_headers",
    srcs = ["parallel_device_lib.h"],
)

filegroup(
    name = "lib_sources",