feat(dflash): support multimodal (VL) training with plain-rope drafts - #1050

feat(dflash): support multimodal (VL) training with plain-rope drafts#1050

curnane-lab wants to merge 3 commits into

vllm-project:mainvllm-project/speculators:mainfrom

curnane-lab:add_vl_dflashcurnane-lab/speculators:add_vl_dflashCopy head branch name to clipboard