* Better **extensibility**.
* etc.

...all this while keeping the **same Python API**. In most of the cases, for simple models, you can simply upgrade the Pydantic version and get all the benefits. 🚀

In some cases, for pure data validation and processing, you can get performance improvements of **20x** or more. This means 2,000% or more. 🤯

* `tf.tpu.experimental.embedding.TPUEmbeddingV2`
    * Add `compute_sparse_core_stats` for sparse core users to profile the  data with this API to get the `max_ids` and `max_unique_ids`. These numbers will be needed to configure the sparse core embedding mid level api.
    * Remove the `preprocess_features` method since that's no longer needed.

## Thanks to our Contributors

NewNat().ExpandFor(m) n := uint(len(rr.limbs)) mLen := uint(m.BitLen()) logR := _W * n // We start by computing R = 2^(_W * n) mod m. We can get pretty close, to // 2^⌊log₂m⌋, by setting the highest bit we can without having to reduce. rr.limbs[n-1] = 1 << ((mLen - 1) % _W) // Then we double until we reach 2^(_W * n). for i := mLen - 1; i < logR; i++ { rr.Add(rr, m) } // Next we need to get from R to 2^(_W * n) R mod m (aka from one to R in // the Montgomery domain, meaning we can use Montgomery multiplication now)....

NewNat().ExpandFor(m) n := uint(len(rr.limbs)) mLen := uint(m.BitLen()) logR := _W * n // We start by computing R = 2^(_W * n) mod m. We can get pretty close, to // 2^⌊log₂m⌋, by setting the highest bit we can without having to reduce. rr.limbs[n-1] = 1 << ((mLen - 1) % _W) // Then we double until we reach 2^(_W * n). for i := mLen - 1; i < logR; i++ { rr.Add(rr, m) } // Next we need to get from R to 2^(_W * n) R mod m (aka from one to R in // the Montgomery domain, meaning we can use Montgomery multiplication now)....