It's debatable. Given that this is basically the only requirement that MIT imposes on reuse, and also the only thing that the original author gets out of it, not doing proper attribution is concerning.
I believe that dropping an email will be enough and that the lack of attribution is not misunderstanding on how licensing works. The level of licensing ignorance in not so old developers is quite shocking. We are no longer in the 90s where most developers were trying to understand the legal terms of the GPL, BSD, MIT, ...
Normally I would agree, but in case of Ollama specifically, even setting aside the license, they've been publicly called out on downplaying the fact that most of their functionality is provided by llama.cpp before many times, but haven't changed their messaging at all, or even acknowledged the call-outs. And, well, this has been an open issue in their public repo for a while now, accumulating comments and upvotes, so surely someone have seen it before. So in this case specifically I'm leaning towards malicious non-compliance, solely because they really want to make the impression that there's more to Ollama than there really is.