frost(sdpa): GQA/MQA, padding mask, right-band widening, and sink-token gradients for the SM120 f16 backward - #557

frost(sdpa): GQA/MQA, padding mask, right-band widening, and sink-token gradients for the SM120 f16 backward#557

vedaanta merged 8 commits into

NVIDIA:developNVIDIA/cudnn-frontend:developfrom

Adnios:feat/sm120_sdpa_bwd_padding_maskAdnios/cudnn-frontend:feat/sm120_sdpa_bwd_padding_maskCopy head branch name to clipboard