AbstractContext* ctx_;
};

// Returns the number of elements in the gradient tensor.
int64_t TapeVSpace::NumElements(AbstractTensorHandle* tensor) const {
  // TODO(srbs): It seems like this is used only for performance optimization
  // and not for correctness. The only downside of keeping this 1 seems to be
  // that the gradient accumulation is unbounded and we will never
  // aggressively aggregate accumulated gradients to recover memory.

        // copies/broadcasts. Other implicit copies may be supported eventually,
        // but need special handling for gradients (gradient of copy-on is not
        // just copy-off but includes a sum) and consideration of performance.
        //
        // TODO(allenl): There may be smarter ways to do this copy in some
        // cases, i.e. with a collective broadcast. We'll need to be careful

// ----------------------------------------------------------------------------
namespace tf_gcs_filesystem {
// TODO(vnvo2409): Use partial reponse for better performance.
// 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.