vllm.attention.backends

Modules:

Name	Description
`abstract`
`differential_flash_attn`	" An implementation of https://arxiv.org/pdf/2410.05258
`dual_chunk_flash_attn`	Attention layer with Dual chunk flash attention and sparse attention.
`flash_attn`	Attention layer with FlashAttention.
`flashmla`
`mla`
`placeholder_attn`
`rocm_aiter_mla`
`rocm_flash_attn`	Attention layer ROCm GPUs.
`triton_mla`
`utils`	Attention backend utils
`xformers`	Attention layer with xFormers and PagedAttention.