sliceForAppend takes a slice and a requested number of bytes. It returns a // slice with the contents of the given slice followed by that many bytes and a // second slice that aliases into it and contains only the extra bytes. If the // original slice has sufficient capacity then no allocation is performed. func sliceForAppend(in []byte, n int) (head, tail []byte) { if total := len(in) + n; cap(in) >= total { head = in[:total] } else { head = make([]byte, total) copy(head, in) } tail = head[len(in):] return...

    *   They are off by default. Enable them by setting the environment variable
        `TF_ENABLE_ONEDNN_OPTS=1`.
    *   We do not recommend using them in GPU systems, as they have not been
        sufficiently tested with GPUs yet.
*   TensorFlow pip packages are now built with CUDA11.2 and cuDNN 8.1.0

## Breaking Changes

*   The `TF_CPP_MIN_VLOG_LEVEL` environment variable has been renamed to