Training and building models is well outside the Pis arena. That requires big processing. The inference engine on the other hand is perfectly possible. It's memory intensive, but the difference between the 4 and 8GB devices is small.Probably true, but I am not sure ML would be considered normal use, which is what the OP wants to be doing. It's not my field of expertise but I've heard people say don't even consider a Pi for training or building models. I am sure I saw something which indicated the Hailo Dataflow Compiler needed 32GB of RAM but I don't know if what the OP is planning is comparable.But probably not noticeable in normal usage.
Statistics: Posted by jamesh — Fri Jul 12, 2024 4:41 pm