Train a LLM from Scratchgithub.com/raiyanyahya 3 pointslinhns4 months ago1 commentSaveHideCopy link On HNComments−subtick4moCurious — how did you handle training stability early on? Was convergence an issue without heavy tuning?
Comments
Curious — how did you handle training stability early on? Was convergence an issue without heavy tuning?