vllm_omni.diffusion.attention.backends.flashinfer_attn ¶

HAS_FLASHINFER `module-attribute` ¶

HAS_FLASHINFER = True

logger = init_logger(__name__)

accept_output_buffer: bool = True

get_impl_cls() -> type[FlashInferAttentionImpl]

get_name() -> str

get_supported_head_sizes() -> list[int]

supports_attention_mask() -> bool

causal = causal

softmax_scale = softmax_scale

forward_cuda(
    query: Tensor,
    key: Tensor,
    value: Tensor,
    attn_metadata: AttentionMetadata | None = None,
) -> Tensor