How do I find the maximum number of tokens for the input area of a model like gpt-j ? I am looking at the model card, but I do not know how to read it properly. I am interested in gpt-j, and I think the size of the input area is either 2048 or 4096. I think when the input gets bigger than this many tokens the input area starts ‘windowing’. Correct me but I think the start of the input is read by the model but is scrolled through so that the end of the input can be read.
I am looking at this page. There is a chart on the left. Maybe someone can point out a description of this kind of chart on another page?