```

...before this change it would raise an error when resetting the context variable, because the `contextvars` context was different, because of the way it was implemented.

**Note**: You probably don't need `contextvars`, and you should probably avoid using them. But they are powerful and useful in some advanced scenarios, for example, migrating from code that used Flask's `g` semi-global variable.

            the first batch we "peek" from the generator. This used to happen
            implicitly in `Model._standardize_user_data`. Long-term, a solution
            where the `DataAdapter` doesn't need to call the Model is probably
            preferable.
    *   The SavedModel format now supports all Keras built-in layers (including
        metrics, preprocessing layers, and stateful RNN layers)

a + b, and returns v. func (v *Element) Add(a, b *Element) *Element { v.l0 = a.l0 + b.l0 v.l1 = a.l1 + b.l1 v.l2 = a.l2 + b.l2 v.l3 = a.l3 + b.l3 v.l4 = a.l4 + b.l4 // Using the generic implementation here is actually faster than the // assembly. Probably because the body of this function is so simple that // the compiler can figure out better optimizations by inlining the carry // propagation. return v.carryPropagateGeneri() } // Subtract sets v = a - b, and returns v. func (v *Element) Subtract(a,...

- Updated the API documentation for Service.Spec.IPFamily to warn that its exact
  semantics will probably change before the dual-stack feature goes GA, and users
  should look at ClusterIP or Endpoints, not IPFamily, to figure out if an existing