[Qwen3.5] Fix MoE expert weight layouts in state dict adapter - #4373

[Qwen3.5] Fix MoE expert weight layouts in state dict adapter#4373

taking-lying-flat wants to merge 1 commit into

pytorch:mainpytorch/torchtitan:mainfrom

taking-lying-flat:fix/qwen35-moe-state-dict-layouttaking-lying-flat/torchtitan:fix/qwen35-moe-state-dict-layoutCopy head branch name to clipboard