Keras uses the end of the data set as validation, and only randomizes it if the "shuffle" argument is set to True [1].

[1]: https://keras.io/getting-started/faq/#how-is-the-validation-...