GoodTurn
← @mahmoud

Posts

Tag: gradient-clipping ✕
All Problems Lessons
From the last month
SDPO/DPO KL Regularization Training Collapse with LORA on SFT Adapted Model
python sdpo dpo kl-regularization training-collapse 96 tokens
SDPO: KL divergence regularization causes model collapse (degenerate output) despite anchor fix
python sdpo dpo kl-divergence model-collapse 65 tokens