*   Add `show_trainable` option to `tf.keras.utils.model_to_dot` to display layer trainable status in model plots.
 *   Add ability to save a `tf.keras.utils.FeatureSpace` object, via `feature_space.save("myfeaturespace.keras")`, and reload it via `feature_space = tf.keras.models.load_model("myfeaturespace.keras")`.
*   Added utility `tf.keras.utils.to_ordinal` to convert class vector to ordinal regression / classification matrix.

* ♻ Change a `dict()` for `{}` in `fastapi/utils.py`. PR [#3138](https://github.com/tiangolo/fastapi/pull/3138) by [@ShahriyarR](https://github.com/ShahriyarR).
* ♻ Move internal variable for errors in `jsonable_encoder` to put related code closer. PR [#4560](https://github.com/tiangolo/fastapi/pull/4560) by [@GuilleQP](https://github.com/GuilleQP).

unrolled and optimized. func counterCrypt(b *aes.Block, out, in []byte, counter *[gcmBlockSize]byte) { enc := aes.EncryptionKeySchedul(b) rounds := len(enc)/4 - 1 counterCryptASM(rounds, out, in, counter, &enc[0]) } // paddedGHASH pads data with zeroes until its length is a multiple of // 16-bytes. It then calculates a new value for hash using the ghash // algorithm. func paddedGHASH(hash *[16]byte, data []byte, productTable *[256]byte) { if siz := len(data) - (len(data) % gcmBlockSize); siz > 0 { gcmHash(hash[:],...

unrolled and optimized. func counterCrypt(b *aes.Block, out, in []byte, counter *[gcmBlockSize]byte) { enc := aes.EncryptionKeySchedul(b) rounds := len(enc)/4 - 1 counterCryptASM(rounds, out, in, counter, &enc[0]) } // paddedGHASH pads data with zeroes until its length is a multiple of // 16-bytes. It then calculates a new value for hash using the ghash // algorithm. func paddedGHASH(hash *[16]byte, data []byte, productTable *[256]byte) { if siz := len(data) - (len(data) % gcmBlockSize); siz > 0 { gcmHash(hash[:],...