Skip to content

Comment on NanoGPT Speedrun Frontier

Comments

Neat!

The graphs show the "best validated result" for each model. I wonder how much variation there is between runs for a model?

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.