is recommended to use the
    [Keras mixed precision API](https://www.tensorflow.org/guide/mixed_precision)
    instead.

*   `tf.lite`:

    *   Remove `experimental.nn.dynamic_rnn`, `experimental.nn.TfLiteRNNCell`
        and `experimental.nn.TfLiteLSTMCell` since they're no longer supported.
        It's recommended to just use

SignExternalMuDeterm(priv, μ) if err != nil { return err } return VerifyExternalMu(priv.PublicKey(), μ, sig) }) } var fipsSelfTest = sync.OnceFunc(func() { fips140.CAST("ML-DSA-44", fips140CAST) }) // fips140CAST covers all rejection sampling paths, as recommended by IG 10.3.A, // and as tested by TestCASTRejectionPat. It tests only one parameter set as // allowed by Note26. It tests the modified version of Algorithm 7 and 8 with a // fixed mu/μ, as allowed by IG 10.3.A, Resolution 15. It compares sk and...

facilitate reviewing the implementation against the NIST FIPS 203 document. // // Reviewers unfamiliar with polynomials or linear algebra might find the // background at https://words.filippo.io/kyber-math/ useful. // // This file implements the recommended parameter set ML-KEM-768. The ML-KEM-1024 // parameter set implementation is auto-generated from this file. // //go:generate go run generate1024.go -input mlkem768.go -output mlkem1024.go import ( "bytes" "crypto/internal/fips140" "crypto/internal/fips140/drbg"...