flashinfer.fused_moe.CudnnFrostMxfp8Config

class flashinfer.fused_moe.CudnnFrostMxfp8Config

Frost MXFP8 MoE on SM107, with packed precomputed routing.

Register prepared weights under cudnn_frost_mxfp8. The canonical cutlass_mxfp8 view is also accepted without conversion. H/I must be multiples of 128. Pass the layer’s quant config to prepare_activations.

__init__() → None

Methods

__init__()
prepare_activations(hidden_states_bf16, *[, ...])	Quantize BF16 activations to MXFP8 for MXFP8×MXFP8.
prepare_weights(w1_bf16, w2_bf16, *, ...[, ...])	Quantize canonical BF16 weights into CUTLASS MXFP8.
supported(arch)