GradientTape<Gradient, BackwardFunction, TapeTensor>* backward_tape,
        bool accumulating)
        : backward_tape(backward_tape), accumulating(accumulating) {}
    // Set temporarily while in the Accumulate method; if backward_tape is not
    // nullptr then we forward op executions to it so Accumulate can compute a
    // backward pass on its backward function.
    //

    and
    [tensorflow/python](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/python).
    TensorFlow has passed version 1.0 and hence cannot make
    non-backward-compatible API changes without a major release. Reviewers of
    your pull request will comment on any API compatibility issues

      with `distributed_save`, it reads the snapshot while it is being written.
      For snapshots written with regular `save`, it waits for the snapshot until
      it's finished. The default is `False` for backward compatibility. Users of
      `distributed_save` are recommended to set it to `True`.

* `tf.tpu.experimental.embedding.TPUEmbeddingV2`
    * Add `compute_sparse_core_stats` for sparse core users to profile the 

            'TF_CUBLAS_VERSION',
            'TF_CUDNN_VERSION',
            'TF_TENSORRT_VERSION',
            'TF_NCCL_VERSION',
            'TF_CUDA_PATHS',
            # Items below are for backwards compatibility when not using
            # TF_CUDA_PATHS.
            'CUDA_TOOLKIT_PATH',
            'CUDNN_INSTALL_PATH',
            'NCCL_INSTALL_PATH',
            'NCCL_HDR_PATH',