On Thu, Mar 4, 2021, 3:30 PM <[email protected]> wrote:

> My code is 109 lines,
>

Which currently does what?

GPT is ~400.

How do you know? The GPT-3 release on GitHub doesn't seem to have source
code. But there is a 3.8 MB download.

But 13,000 lines, bro, what are you doing in the code???

Probably a quarter of that is data compression and the rest is file
management, encryption, threading, and other features. Zpaq embeds the
decompression algorithm in the archive in a sandboxed virtual machine
language called ZPAQL. The decompressor translates it into x86 or
interprets it. There are ZPAQL models implementing LZ77, BWT, and PAQ style
context mixing. It also implements dedupe by matching SHA1 hashes of file
fragments split on content dependent boundaries. It's primary use is for
append-only incremental backups that can be rolled back to earlier versions.

Anyway the various PAQ versions that only do compression are a few thousand
lines. There's lots of code to handle special cases like JPEG, x86, BMP,
WAV, tables, databases, text, numeric data, etc.

>
>

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Tef43964b9365fffb-M0f9c47d6d6f179000dea6774
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to