`float16` for acceleration on Volta and Turing Tensor Cores. This feature
    can be enabled by wrapping an optimizer class with
    `tf.train.experimental.enable_mixed_precision_graph_rewrite()`.
*   Add environment variable `TF_CUDNN_DETERMINISTIC`. Setting to "true" or "1"
    forces the selection of deterministic cuDNN convolution and max-pooling

  if (op_reg_data->shape_inference_fn == nullptr) {
    status->status =
        InvalidArgument("No shape inference function exists for op '",
                        node_def.op(), "', did you forget to define it?");
    return;
  }

  status->status = c.Run(op_reg_data->shape_inference_fn);
  if (!status->status.ok()) return;

  // Set output_shapes.
  TF_ShapeAndTypeList* output_shapes_result =