Before Starlette `0.24.0`, a new instance of each middleware class would be created when a new middleware was added. That normally was not a problem, unless the middleware class expected to be created only once, with only one instance, that happened in some cases. This upgrade would solve those cases (thanks [@adriangb](https://github.com/adriangb)! Starlette PR [#2017](https://github.com/encode/starlette/pull/2017)). Now the middleware class instances are created once, right before the first request...

        not supported when used with `tf.distribute.Strategy` outside of
        built-in training loops like `tf.keras` `compile`/`fit`.
    *   Wraps losses passed to the `compile` API (strings and v1 losses) which
        are not instances of v2 `Loss` class in `LossWrapper` class. => All
        losses will now use `SUM_OVER_BATCH_SIZE` reduction as default.

New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point()} table[0].Set(q) for i := 1; i < 15; i += 2 { table[i].Double(table[i/2]) table[i+1].Add(table[i], q) } // Instead of doing the classic double-and-add chain, we do it with a // four-bit window: we double four times, and then add [0-15]P. t := New{{.P}}Point() p.Set(New{{.P}}Point()) for i, byte := range scalar { // No need to double on the first iteration, as p is the identity at...

New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point(), New{{.P}}Point()} table[0].Set(q) for i := 1; i < 15; i += 2 { table[i].Double(table[i/2]) table[i+1].Add(table[i], q) } // Instead of doing the classic double-and-add chain, we do it with a // four-bit window: we double four times, and then add [0-15]P. t := New{{.P}}Point() p.Set(New{{.P}}Point()) for i, byte := range scalar { // No need to double on the first iteration, as p is the identity at...