// destination tensor to be placed in host memory).
// If async execution is enabled, the copy may be enqueued and the call will
// return "non-ready" handle. Else, this function returns after the copy has
// been done.
TF_CAPI_EXPORT extern TFE_TensorHandle* TFE_TensorHandleCopyToDevice(
    TFE_TensorHandle* h, TFE_Context* ctx, const char* device_name,
    TF_Status* status);

// Debugging/Profiling information for TFE_TensorHandle

      # it, and print that list for debugging. Not really clear if this is
      # helpful since the only examples I've seen are enormous.
      bazel query "rdeps(kind(py_test, $(cat $BATS_TEST_TMPDIR/deps)), $dep, 1)"
    done < $BATS_TEST_TMPDIR/missing_deps
    exit 1
  fi
}

# The Python package is not allowed to depend on any CUDA packages.
@test "Pip package doesn't depend on CUDA" {
  bazel cquery \

Use `clang-tidy` to check your C/C++ changes. To install `clang-tidy` on ubuntu:16.04, do:

```bash
apt-get install -y clang-tidy
```

You can check a C/C++ file by doing:


```bash
clang-format <my_cc_file> --style=google > /tmp/my_cc_file.cc
diff <my_cc_file> /tmp/my_cc_file.cc
```

#### Python coding style

Changes to TensorFlow Python code should conform to

  // Set device to CPU since there is no version of split for int32 on GPU
  // TODO(iga): Convert all these helpers and tests to use floats because
  // they are usually available on GPUs. After doing this, remove TF_SetDevice
  // call in c_api_function_test.cc
  TF_SetDevice(desc, "/cpu:0");
  *op = TF_FinishOperation(desc, s);
  ASSERT_EQ(TF_OK, TF_GetCode(s)) << TF_Message(s);
  ASSERT_NE(*op, nullptr);
}

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

namespace tensorflow {

// Checks that all schemes provided by a plugin are valid.
// TODO(b/139060984): More validation could be done here, based on supported
// charset, maximum length, etc. Punting it for later.
static Status ValidateScheme(const char* scheme) {
  if (scheme == nullptr)
    return errors::InvalidArgument(

// TODO(vnvo2409): We could do some cleanups like `return TF_SetStatus`.
// TODO(vnvo2409): Refactor the filesystem implementation when
// https://github.com/googleapis/google-cloud-cpp/issues/4482 is done.
GCSFile::GCSFile(google::cloud::storage::Client&& gcs_client)
    : gcs_client(gcs_client), block_cache_lock() {
  const char* append_mode = std::getenv(kAppendMode);

// 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
// Tensors need gradients to be computed (Tensors which are not used do not need

  auto collective_executor_handle = context->GetCollectiveExecutorHandle();
  tensorflow::Notification done;
  collective_executor_handle->get()->remote_access()->CheckPeerHealth(
      task, timeout_in_ms, [&done, status](const Status& s) {
        status->status = s;
        done.Notify();
      });
  done.WaitForNotification();
}

TF_ShapeAndTypeList* TF_NewShapeAndTypeList(int num_items) {

      # it, and print that list for debugging. Not really clear if this is
      # helpful since the only examples I've seen are enormous.
      bazel cquery "rdeps(kind(py_test, $(cat $BATS_TEST_TMPDIR/deps)), $dep, 1)"
    done < $BATS_TEST_TMPDIR/missing_deps
    exit 1
  fi
}

# The Python package is not allowed to depend on any CUDA packages.
@test "Pip package doesn't depend on CUDA" {
  bazel cquery \

}

TEST_P(CppGradients, TestMatMulGrad) {
  // TODO(vnvo2409): Figure out why `gradient_checker` does not work very
  // well with `MatMul` and remove `TestMatMul*` in
  // `mnist_gradients_test` when done.
  GTEST_SKIP();

  float A_vals[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f};
  int64_t A_dims[] = {3, 3};
  AbstractTensorHandlePtr A;
  {
    AbstractTensorHandle* A_raw;