all transformations.
    *   The optimization `parallel_batch` now becomes default if not disabled by
        users, which will parallelize copying of batch elements.
    *   Added the ability for `TensorSliceDataset` to identify and handle inputs
        that are files. This enables creating hermetic SavedModels when using
        datasets created from files.

result // into dst. cnt is the initial count value and will be updated with the next // count value. The length of dst must be greater than or equal to the length // of src. func counterCrypt(g *GCM, dst, src []byte, cnt *[gcmBlockSize]byte) { // Copying src into a buffer improves performance on some models when // src and dst point to the same underlying array. We also need a // buffer for counter values. var ctrbuf, srcbuf [2048]byte for len(src) >= 16 { siz := len(src) if len(src) > len(ctrbuf)...