same implementation as their `tf.losses` equivalent.
*   For Keras model, the individual call of `Model.evaluate` uses no cached data
    for evaluation, while `Model.fit` uses cached data when `validation_data`
    arg is provided for better performance.
*   Adds a `save_traces` argument to `model.save`/ `tf.keras.models.save_model`

- Avoid caching the Azure VMSS instances whose network profile is nil ([#100948](https://github.com/kubernetes/kubernetes/pull/100948), [@feiskyer](https://github.com/feiskyer)) [SIG Cloud Provider]
- Azure: avoid setting cached Sku when updating VMSS and VMSS instances ([#102005](https://github.com/kubernetes/kubernetes/pull/102005), [@feiskyer](https://github.com/feiskyer)) [SIG Cloud Provider]

// Add sets v = p + q, and returns v. func (v *Point) Add(p, q *Point) *Point { checkInitialized(p, q) qCached := new(projCached).FromP3(q) result := new(projP1xP1).Add(p, qCached) return v.fromP1xP1(result) } // Subtract sets v = p - q, and returns v. func (v *Point) Subtract(p, q *Point) *Point { checkInitialized(p, q) qCached := new(projCached).FromP3(q) result := new(projP1xP1).Sub(p, qCached) return v.fromP1xP1(result) } func (v *projP1xP1) Add(p *Point, q *projCached) *projP1xP1 { var YplusX,...

* ♻ Strip empty whitespace from description extracted from docstrings. PR [#2821](https://github.com/tiangolo/fastapi/pull/2821) by [@and-semakin](https://github.com/and-semakin).
* 🐛 Fix cached dependencies when using a dependency in `Security()` and other places (e.g. `Depends()`) with different OAuth2 scopes. PR [#2945](https://github.com/tiangolo/fastapi/pull/2945) by [@laggardkernel](https://github.com/laggardkernel).