*   Unary GPU kernels: Abs, Atanh, Acos, Acosh, Asin, Asinh, Atan, Cos, Cosh, Sin, Sinh, Tan, Tanh.
    *   Binary GPU kernels: AddV2, Sub, Div, DivNoNan, Mul, MulNoNan, FloorDiv, Equal, NotEqual, Greater, GreaterEqual, LessEqual, Less.

* `tf.lite`
    * Add experimental supports conversion of models that may be larger than 2GB before buffer deduplication

  EXPECT_TRUE(absl::StrContains(s, expected))
      << "'" << s << "' does not contain '" << expected << "'";
}

// Returns the GPU device name if there is one (with arbitrary tie breaking if
// there are more than one), or "" otherwise.
string GPUDeviceName(TF_Session* session) {
  std::unique_ptr<TF_Status, decltype(&TF_DeleteStatus)> status(
      TF_NewStatus(), TF_DeleteStatus);
  TF_Status* s = status.get();

                                     " but actual rank is ", rank);
    return;
  }

  if (num_dims == 0) {
    // Output shape is a scalar.
    return;
  }

  // Rank is greater than 0, so fill in the values, if known, and
  // -1 for unknown values.
  for (int i = 0; i < num_dims; ++i) {
    auto dim = ic->Dim(shape, i);
    int64_t value = -1;
    if (ic->ValueKnown(dim)) {

      auto& indices = iter->second;
      if (std::find(indices.begin(), indices.end(), idx) != indices.end()) {
        return InvalidArgument("TF_Output ", node->name(), ":", idx,
                               " appears more than once in the input list");
      }
      indices.push_back(idx);
    }
  }
  return absl::OkStatus();
}

// Converts `noutputs` and `outputs` into `outputs_tensors` and does various
// checks while doing so.

      const std::function<void(BackwardFunction*)>& backward_function_deleter);

  void DeleteTrace(int64_t tensor_id);

  // Consumes the internal state of the tape (so cannot be called more than
  // once) and produces the gradient of the target tensors with respect to the
  // source tensors. The output gradients are used if not empty and not
  // null. The result is populated with one tensor per target element.

            all_valid = False
          if ver < 35:
            print('WARNING: XLA does not support CUDA compute capabilities '
                  'lower than sm_35. Disable XLA when running on older GPUs.')
      else:
        ver = float(m.group(0))
        if ver < 3.0:
          print('ERROR: TensorFlow only supports CUDA compute capabilities 3.0 '

# TODO(rostam): Remove when `cc_shared_library` is enabled by default
build --experimental_cc_shared_library

# cc_shared_library ensures no library is linked statically more than once.
build --experimental_link_static_libraries_once=false

# Prevent regressions on those two incompatible changes
# TODO: remove those flags when they are flipped in the default Bazel version TF uses.