# package
  bazel cquery --keep_going 'deps(//tensorflow/tools/pip_package:wheel)' | sort -u > $BATS_TEST_TMPDIR/pip_deps
  # Find all Python py_test targets not tagged "no_pip" or "manual", excluding
  # any targets in ignored packages. Combine this list of targets into a bazel
  # query list (e.g. the list becomes "target+target2+target3")

    // These XLA flags are needed to trigger XLA properly from C (more generally
    // non-Python) clients. If this API is called again with `enable` set to
    // false, it is safe to keep these flag values as is.
    tensorflow::MarkForCompilationPassFlags* flags =
        tensorflow::GetMarkForCompilationPassFlags();
    flags->tf_xla_cpu_global_jit = true;
    flags->tf_xla_min_cluster_size = 1;
  } else {

build:release_cpu_linux_base --repo_env=CC="/usr/lib/llvm-17/bin/clang"
build:release_cpu_linux_base --repo_env=BAZEL_COMPILER="/usr/lib/llvm-17/bin/clang"
# Test-related settings below this point.
test:release_linux_base --build_tests_only --keep_going --test_output=errors --verbose_failures=true
test:release_linux_base --local_test_jobs=HOST_CPUS
test:release_linux_base --test_env=LD_LIBRARY_PATH
# Give only the list of failed tests at the end of the log

template <typename Gradient, typename BackwardFunction, typename TapeTensor>
class GradientTape {
 public:
  // If `persistent` is true, GradientTape will not eagerly delete backward
  // functions (and hence the tensors they keep alive). Instead, everything
  // is deleted in ~GradientTape. Persistent GradientTapes are useful when
  // users want to compute multiple gradients over the same tape.

# 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
# -x: log all commands

# (error on undefined variables). This ensures that (a) every TFCI variable
# has an explicit default value, and (b) no script can accidentally use a
# variable that doesn't exist. Please keep this list in alphabetical order.
#
# Find usage in scripts with e.g.:
#   cd ci/official
#   ls *.sh utilities/*.sh | xargs grep -H --color=always TFCI_ARG_HERE
# You may also get an overview, e.g.:
#   cd ci/official

    sum_dims.reset(sum_dims_raw);
  }

  // Reduce sum the output on all dimensions.
  TF_RETURN_IF_ERROR(ops::Sum(ctx, model_out.get(), sum_dims.get(), &outputs[0],
                              /*keep_dims=*/false, "sum_output"));
  return absl::OkStatus();
}
// ========================= End Helper Functions==============================

Status CalcNumericalGrad(AbstractContext* ctx, Model forward,

  //   Outputs:
  std::vector<TensorHandlePtr> op_outputs_ TF_GUARDED_BY(execution_mutex_);
  // TF_Status is an incomplete type and so can't be stack allocated. To avoid
  // unnecessary allocations each Execute call, we keep one heap-allocated
  // version for the thread.
  StatusPtr status_ TF_GUARDED_BY(execution_mutex_);

  const std::string device_;
  ExecutorPtr executor_ TF_GUARDED_BY(execution_mutex_);