# Don't fail if these modules are not available.
# For e.g. this file will be originally placed under tensorflow/_api/v1 which
# does not have "python", "core" directories. Then, it will be copied
# to tensorflow/ which does have these two directories.
try:
  del python
except NameError:
  pass
try:
  del core
except NameError:
  pass
try:
  del compiler
except NameError:

# Don't fail if these modules are not available.
# For e.g. this file will be originally placed under tensorflow/_api/v1 which
# does not have "python", "core" directories. Then, it will be copied
# to tensorflow/ which does have these two directories.

try:
  del python
except NameError:
  pass
try:
  del core
except NameError:
  pass
try:
  del compiler
except NameError:

  *input = {oper, 0};
  return true;
}

// Copies `src_graph` into `dst_graph`. Any node in `src_graph` with input
// `src_inputs[i]` will have that input replaced with `dst_inputs[i]`.  `prefix`
// will be prepended to copied node names. `control_deps` are nodes in
// `dst_graph` that the copied `src_graph` nodes will have control dependencies

// backwards functions are deleted.
// If `persistent_tape` is false, op_tape is cleared and backwards functions
// not needed for gradient computation are deleted. Backwards functions that
// are needed, are copied and returned in BackpropInitialState.
template <typename BackwardFunction, typename TapeTensor>
BackpropInitialState<BackwardFunction, TapeTensor> PrepareBackprop(

TF_CAPI_EXPORT extern bool TFE_ExecutorIsAsync(TFE_Executor*);

// Causes the calling thread to block till all ops dispatched in this executor
// have been executed. Note that "execution" here refers to kernel execution /
// scheduling of copies, etc. Similar to sync execution, it doesn't guarantee
// that lower level device queues (like GPU streams) have been flushed.
//
// This call may not block for execution of ops enqueued concurrently with this
// call.

                absl::Span<AbstractTensorHandle*> outputs, bool use_function) {
  if (use_function) {
    const char* fn_name = "test_fn";
    core::RefCountPtr<AbstractFunction> scoped_func;
    // Returning null tensors from a tf.function is not supported, so we keep
    // track of indices in the model's outputs are nullptr in this set.
    // The FunctionDef only outputs the non-null tensors. We later pad the

*   TF Core:
    *   `tf.constant` always creates CPU tensors irrespective of the current
        device context.
    *   Eager `TensorHandles` maintain a list of mirrors for any copies to local
        or remote devices. This avoids any redundant copies due to op execution.
    *   For `tf.Tensor` & `tf.Variable`, `.experimental_ref()` is no longer
        experimental and is available as simply `.ref()`.

  auto* rewrite_options =
      config->mutable_graph_options()->mutable_rewrite_options();
  rewrite_options->set_scoped_allocator_optimization(
      tensorflow::RewriterConfig::ON);
  rewrite_options->mutable_scoped_allocator_opts()->add_enable_op(
      "CollectiveReduce");

  if ((tensorflow::IsGoogleCudaEnabled() || tensorflow::IsBuiltWithROCm()) &&
      num_virtual_gpus > 0) {

# with clang. It does not imply that CUDA support has been enabled.
alias(
    name = "is_cuda_compiler_clang",
    actual = if_oss(
        "@local_config_cuda//:is_cuda_compiler_clang",
        "@local_config_cuda//cuda:TRUE",
    ),
)

# Config setting that is satisfied when CUDA device code should be compiled

# CROSS-COMPILE ARM64 PYCPP
build:cross_compile_linux_arm64_pycpp_test --config=linux_arm64_pycpp_test
# Tests that fail only when cross-compiled
build:cross_compile_linux_arm64_pycpp_test -//tensorflow/compiler/mlir/quantization/stablehlo:convert_tf_quant_to_mhlo_int_test
# MACOS ARM64 PYCPP