Building signatures from these allows TF Serving to honor requests using the
    different APIs (Classify, Regress, and Predict). Furthermore,
    `serving_input_receiver_fn()` may now specify alternative subsets of nodes
    that may act as inputs. This allows, for instance, producing a prediction
    signature for a classifier that accepts raw `Tensors` instead of a
    serialized `tf.Example`.

	SUBSD X11, X11                          // f2450f5cdb
	SUBSS (BX), X2                          // f30f5c13
	SUBSS (R11), X2                         // f3410f5c13
	SUBSS X2, X2                            // f30f5cd2
	SUBSS X11, X2                           // f3410f5cd3
	SUBSS (BX), X11                         // f3440f5c1b
	SUBSS (R11), X11                        // f3450f5c1b
	SUBSS X2, X11                           // f3440f5cda

### Feature

- AWS cloudprovider supports auto-discovering subnets without any kubernetes.io/cluster/<clusterName> tags. It also supports additional service annotation service.beta.kubernetes.io/aws-load-balancer-subnets to manually configure the subnets. ([#97431](https://github.com/kubernetes/kubernetes/pull/97431), [@kishorj](https://github.com/kishorj)) [SIG Cloud Provider]

    * This helps to more easily solve use cases like:
        * Reading a body before and/or after a request (equivalent to a middleware).
        * Run middleware-like code only for a subset of *path operations*.
        * Process a request before passing it to a *path operation function*. E.g. decompressing, deserializing, etc.