You might be interested in the world's smallest data compressor, smile.com. The executable (16 bit DOS) is 256 bytes. It uses a move to front algorithm with Elias coding of the queue position. It compresses enwik9 only slightly to 695 MB. http://mattmahoney.net/dc/text.html#6955
Anyway, GPT-2 is considerably larger if you count the tensorflow import that does most of the work. As far as I can tell, it's a couple thousand source code files on the order of 1M lines. On Fri, Mar 5, 2021, 2:17 AM <[email protected]> wrote: > My 109 lines of code compresses 1,000,000 bytes to 251,029 bytes, or > ~21.4MB from 100MB, per comparison to other Hutter Prize scores. Almost > ready to show it. It's mostly the same code but smaller. I just started > back ~13 days ago. > > GPT-2 code is on their website Matt....yes it's around 400 lines. It is > even on my computer :-) i'm never gonna let it go either. > > The reason there is so many reflexes like thumb-index finger connection/ > sneeze/ etc that evolution learnt to force install is because it had no > other way but to hardcode so much. GPT is only ~400 lines of code, and we > saw what that can do. > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/Tef43964b9365fffb-M58a04102b6203b154baab442> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tef43964b9365fffb-Mcd7211f13531579900e71bca Delivery options: https://agi.topicbox.com/groups/agi/subscription
