nanochat

fix scaling laws scripts after the bigram embeddings were removed

submit new time to GPT-2 leaderboard entry: 99 minutes

Autoresearch round 2: smear, backout, and hyperparameter tuning

new leaderboard entry coming from improvements of autoresearch round 1, time to gpt-2 from 2.02 hours to 1.80 hours

All of these improvements were developed by Claude running autonomously over ~2 days using autoresearch. I didn't touch anything - incredible. All tuning was done on d12 but generalized easily to larger models (e.g. d24 in particular). This means we will also get a new "Time to GPT-2" Leaderboard entry, which I will push separately.