
Phi 3 and Arctic: Outlier LMs are hints
Interconnects
Introduction
This chapter delves into the unique features and training techniques of the new Open models, Phi 3 from Microsoft and Arctic from Snowflake, highlighting Arctic's sparse mixture of experts architecture and its coding and reasoning niche. It also discusses the design choices, parameters, experts, and intelligence improvements in the Arctic model for efficient training and inference.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.