[Train] Reduce padding-free embedding output memory - #9893

[Train] Reduce padding-free embedding output memory#9893

taking-lying-flat wants to merge 1 commit into

modelscope:mainmodelscope/ms-swift:mainfrom

taking-lying-flat:agent/fix-qwen3-vl-embedding-memorytaking-lying-flat/ms-swift:agent/fix-qwen3-vl-embedding-memoryCopy head branch name to clipboard