Comment on Train a LLM from ScratchComments−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?