### `tf.keras`:

*   Improvements from the Functional API refactoring:
    *   Functional model construction does not need to maintain a global
        workspace graph, removing memory leaks especially when building many
        models or very large models.
    *   Functional model construction should be ~8-10% faster on average.
    *   Functional models can now contain non-symbolic values in their call

skDecode to // > return values that are not in the correct range. Hence, skDecode // > should only be run on inputs that come from trusted sources. // // so it sounds like it doesn't even want us to check the coefficients are // within bounds, but especially if using this format for key exchange, that // sounds like a bad idea. So we check everything. t1 := make([][n]uint16, k, maxK) for i := range k { tHat := priv.s2[i] for j := range l { tHat = polyAdd(tHat, nttMul(A[i*l+j], priv.s1[j])) } t := inverseNTT(tHat)...