I closed it, with a comment. It appears to be asking only about how keras works. If they edit to clarify a machine learning question, I'll be happy to reopen.
For future reference, you are not obliged to hold off on answering, just because a question may be off topic. I have answered questions that I voted to close, or to migrate to SO, for example. I believe I have done this even since I have had the ability to close instantly (i.e., as a moderator, rather than sending it into the review queue). If you think you can help the OP in the interim, there's no reason not to. (You can also still comment to provide help, if you want—I've done that too.) If you are worried about losing reputation in exchange for your efforts, be aware that we don't delete such threads just because they've been closed, and the roomba won't delete them if there's an answer with a positive score.