flashinfer.fused_moe.BackendOptions¶

Ordered list of backend candidates for dispatch and autotuning.

Each backend config implements supported(arch), where arch uses the CUDA compute-capability encoding documented by valid_for().

__init__(candidates: Tuple[CakeWarpDecodeConfig | PrimsTsConfig | TrtllmFp4Config | TrtllmFp8BlockConfig | TrtllmFp8PerTensorConfig | TrtllmBf16Config | TrtllmMxInt4Config | CutlassBf16Config | CuTileBf16Config | CuTileFp8PerTensorBf16Config | CuTileFp8PerTensorConfig | CuTileMxfp4Bf16Config | CuTileMxfp4Config | CuTileMxfp4Mxfp8Config | CuTileMxfp8Bf16Config | CuTileMxfp8Config | CuTileNvfp4Bf16Config | CuTileNvfp4Config | CutlassW4A16Config | CutlassNvfp4Config | CutlassFp8PerTensorConfig | CutlassFp8BlockConfig | CutlassMxfp8Mxfp4Config | CutlassMxfp8Config | CutlassW4A8Config | CutlassHummingConfig | CuteDslConfig | SM12xFp8Config | SM12xMxfp8Mxfp4Config | SM12xNvfp4Bf16Config | SM12xNvfp4Config | B12xNvfp4Config | B12xW4A16Config, ...] = ()) → None¶

Methods

`__init__`([candidates])
`valid_for`(arch)	Return candidates whose hardware preconditions are met.

Attributes

candidates