replicas taking part in sync training.
    *   Performance improvements for GPU multi-worker distributed training using
        `tf.distribute.experimental.MultiWorkerMirroredStrategy`
    *   Update NVIDIA `NCCL` to `2.5.7-1` for better performance and performance
        tuning. Please see

    * This results in 2x (or more) performance increase for JSON responses.
    * New docs: [Custom Response - JSON Performance](https://fastapi.tiangolo.com/advanced/custom-response/#json-performance).

## 0.129.2 (2026-02-21)

### Internal

- Kubelet: improve the performance when waiting for a synchronization of the node list with the kube-apiserver ([#99336](https://github.com/kubernetes/kubernetes/pull/99336), [@neolit123](https://github.com/neolit123)) [SIG Node]

(2^256-1) else (x1 & (2^256-1)) - 2^256 package edwards25519 import "math/bits" type fiatScalarUint1 uint64 // We use uint64 instead of a more narrow type for performance reasons; see https://github.com/mit-plv/fiat-crypto/pull/1006#issuecomment-892625927 type fiatScalarInt1 int64 // We use uint64 instead of a more narrow type for performance reasons; see https://github.com/mit-plv/fiat-crypto/pull/1006#issuecomment-892625927 // The type fiatScalarMontgomery is a field element in the Montgomery domain....

(2^256-1) else (x1 & (2^256-1)) - 2^256 package edwards25519 import "math/bits" type fiatScalarUint1 uint64 // We use uint64 instead of a more narrow type for performance reasons; see https://github.com/mit-plv/fiat-crypto/pull/1006#issuecomment-892625927 type fiatScalarInt1 int64 // We use uint64 instead of a more narrow type for performance reasons; see https://github.com/mit-plv/fiat-crypto/pull/1006#issuecomment-892625927 // The type fiatScalarMontgomery is a field element in the Montgomery domain....