Train a LLM from Scratch (opens in new tab)

(github.com)

2 pointslinhns5d ago1 comments

1 comments

Curious — how did you handle training stability early on? Was convergence an issue without heavy tuning?

j / k navigate · click thread line to collapse

Curious — how did you handle training stability early on? Was convergence an issue without heavy tuning?

j / k navigate · click thread line to collapse