flashinfer.fused_moe.CuTileMxfp4Bf16Config

class flashinfer.fused_moe.CuTileMxfp4Bf16Config

cuTile MXFP4-weight x BF16-activation backend.

Uses the same prepared weight view as CuTileMxfp4Config. Expert parallelism and fused shared experts are not supported.

__init__() → None

Methods

__init__()
prepare_weights(w1_fp4, w1_block_scale, ...)	Build the shared cutile_mxfp4 weight view.
supported(arch)