I just told it to
Make a cheatsheet with two columns: on the left, Markdown syntax; on the right, the ReStructured Text equivalent.
and it started well but after a minute it seemed to enter a loop outputting the same thing over and over again.
Is this something one should report, or is it something that sometimes happens with ai chats?
I don’t think you need to report this. As far as I know HF Chat is providing relatively direct access to the models. It’s more to show the available models and their inference API than a 100% polished product. And some models aren’t that good, or start to repeat or produce garbage under certain conditions. And Nous-Hermes Mixtral isn’t a model that was made by HF anyways.
Thank you!