please use `optimizer._learning_rate`.
    *   **You implemented a custom optimizer based on the old optimizer.**
        Please set your optimizer to subclass
        `tf.keras.optimizer.legacy.XXX`. If you want to migrate to the new
        optimizer and find it does not support your optimizer, please file
        an issue in the Keras GitHub repo.

avoids/fixes a potential security issue: as the returned object is passed directly to Pydantic, if the returned object was a subclass of the `response_model` (e.g. you return a `UserInDB` that inherits from `User` but contains extra fields, like `hashed_password`, and `User` is used in the `response_model`), it would still pass the validation (because `UserInDB` is a subclass of `User`) and the object would be returned as-is, including the `hashed_password`. To fix this, the declared `response_model`...