NanoPhi, emulating some Of Phi-1.5's success with tinier LM's

2 points by senseiV ↗ HN
i'm trying to replicate at least some of the success of Phi 1.5 on a model 4x smaller, gpt-2 355m.

I have model tasks with examples, it can describe, roleplay, talk about text, make a textbook, QA, simple logic, etc.

Its also uncensored, if you're looking for that

There's a simple github with roadmap, https://github.com/VatsaDev/NanoPhi, check it out there!

0 comments

[ 3.1 ms ] story [ 9.4 ms ] thread

No comments yet.