I recently started out journey with GPT-2 and tried to collect some data for training set. Unfortunately, so far I found less than ten online books with quality that wouldn’t shoot me in the foot, and I’m looking for datasets with good quality.
Already tried to search for it on this site and others, but I’m probably not using the right keywords (text-generation only netted me some dialogue ones so far). Any pointers to where search/look at is appreciated.
Also, can someone give me a rough idea how much data should I collect for a decent training set? Are we talking about millions of words or GBs of text files?