vllm.attention.ops

Modules:

Name	Description
`blocksparse_attention`
`chunked_prefill_paged_decode`
`flashmla`
`hpu_paged_attn`
`ipex_attn`
`merge_attn_states`
`nki_flash_attn`
`paged_attn`
`pallas_kv_cache_update`
`prefix_prefill`
`rocm_aiter_mla`
`rocm_aiter_paged_attn`
`triton_decode_attention`	Memory-efficient attention for decoding.
`triton_flash_attention`	Fused Attention
`triton_merge_attn_states`
`triton_unified_attention`