Is it possible that this is a small model of GPT-3.5?

#6
by Trangle - opened

Considering the almost significant differences in the OpenAI c100 vocabulary, model settings, and performance!

Microsoft org

While some of the design choices are geared towards faster inference via lower kv cache footprints, this is very much a phi-3 small model :)

bapatra changed discussion status to closed

Sign up or log in to comment