Home
last modified time | relevance | path

Searched defs:num_m_block (Results 1 – 2 of 2) sorted by relevance

/external/pytorch/aten/src/ATen/native/transformers/cuda/flash_attn/
Dflash_fwd_launch_template.h66 const int num_m_block = (params.seqlen_q + Kernel_traits::kBlockM - 1) / Kernel_traits::kBlockM; in run_flash_fwd() local
107 const int num_m_block = (params.seqlen_q + Kernel_traits::kBlockM - 1) / Kernel_traits::kBlockM; in run_flash_splitkv_fwd() local
Dflash_bwd_launch_template.h75 const int num_m_block = (params.seqlen_q + Kernel_traits::kBlockM - 1) / Kernel_traits::kBlockM; in run_flash_bwd_seqk_parallel() local