We'd love to accept your patches! Before we can take them, we have to jump a couple of legal hurdles.

Please fill out either the individual or corporate Contributor License Agreement (CLA).

  * If you are an individual writing original source code and you're sure you own the intellectual property, then you'll need to sign an [individual CLA](https://code.google.com/legal/individual-cla-v1.0.html).

  }

  dlm_tensor->dl_tensor.shape = shape_arr->data();
  // There are two ways to represent compact row-major data
  // 1) nullptr indicates tensor is compact and row-majored.
  // 2) fill in the strides array as the real case for compact row-major data.
  // Here we choose option 2, since some frameworks didn't handle the strides
  // argument properly.
  dlm_tensor->dl_tensor.strides = stride_arr->data();

#include "tensorflow/core/platform/errors.h"

namespace tensorflow {
namespace gradients {
namespace {

// TODO(b/172558015): Using the pointer address as the identifier for the tensor
// may lead to collisions. Introduce another way to get a unique id for this
// tensor.
int64_t ToId(const AbstractTensorHandle* t) {
  return static_cast<int64_t>(reinterpret_cast<uintptr_t>(t));
}

// leader and become accessible to all workers in the cluster.
// Currently, a config key can only be set with one value, and subsequently
// setting the same key will lead to errors.
//
// Note that the key-values are only expected to be used for cluster
// configuration data, and should not be used for storing a large amount of data
// or being accessed very frequently.

        convolution algorithms, a regression that was introduced in v2.5. Note
        that nondeterministic out-of-memory events while selecting algorithms
        could still lead to nondeterminism, although this is very unlikely. This
        additional, unlikely source will be eliminated in a later version.
    *   Add deterministic GPU implementations of:

# On linux, we dynamically link small amount of kernels
build:linux --config=dynamic_kernels

# Make sure to include as little of windows.h as possible
build:windows --copt=-DWIN32_LEAN_AND_MEAN
build:windows --host_copt=-DWIN32_LEAN_AND_MEAN
build:windows --copt=-DNOGDI
build:windows --host_copt=-DNOGDI

# MSVC (Windows): Standards-conformant preprocessor mode

    linkstatic = False,
    tags = [
        "manual",
        "nobuilder",
        "notap",
    ],
    visibility = ["//visibility:public"],
    deps = [":windows_filesystem_impl"],
)

# The real implementation of the filesystem.
cc_library(
    name = "windows_filesystem_impl",
    srcs = ["windows_filesystem.cc"],
    copts = get_win_copts(),
    tags = [
        "manual",
        "nobuilder",

    tensorflow::unwrap(op_.get())->SetStepId(step_id_.value());
  }

  TFE_Execute(op_.get(), unwrapped_results.data(), &real_num_outputs, status);
  if (TF_GetCode(status) != TF_OK) {
    cancellation_manager_->StartCancel();
    return;
  }
  unwrapped_results.resize(real_num_outputs);
  outputs->reserve(real_num_outputs);
  for (TFE_TensorHandle* unwrapped_result : unwrapped_results) {
    outputs->emplace_back(unwrapped_result);

# functions (except for tfrun); use dedicated scripts instead and reference them
# specifically. Use your best judgment to keep the scripts in this directory
# lean and easy to follow. When in doubt, remember that for CI scripts, "keep it
# simple" is MUCH more important than "don't repeat yourself."

# -e: abort script if one command fails
# -u: error if undefined variable used