yes this sounds like some variant of a memory error. are you using the OTU-clustered sequences to train your classifier or the raw reads? Use the former, the raw reads will eat up extraordinary amounts of memory.
There is a pre-trained UNITE classifier trained with q2-feature-classifier 2018.11 halfway down the page here. Not exactly what you are looking for, but it would be compatible with an earlier release of QIIME 2.
I hope that helps!