submission 665856

import torch
from task import input_t, output_t

import aiter
from aiter import ActivationType, QuantType
from aiter.fused_moe import fused_moe


@torch.inference_mode()
def custom_kernel(data: input_t) -> output_t:
    """
    与 reference 一致：只走 fused_moe，避免 fmoe_g1u1 触发 module_moe_asm 的长时间 JIT，
    且勿在 import aiter 前设置 AITER_USE_OPUS_MOE_SORTING（会改变排序与编译路径）。
    """
    (
        hidden_states,
        _gw,
        _dw,
        _gws,
        _dws,
        gate_up_weight_shuffled,
        down_weight_shuffled,
        gate_up_weight_scale_shuffled,
        down_weight_scale_shuffled,
        topk_weights,
        topk_ids,
        config,
    ) = data

    hidden_pad = config["d_hidden_pad"] - config["d_hidden"]
    intermediate_pad = config["d_expert_pad"] - config["d_expert"]

    return fused_moe(
        hidden_states,
        gate_up_weight_shuffled,
        down_weight_shuffled,
        topk_weights,
        topk_ids,
        expert_mask=None,
        activation=ActivationType.Silu,
        quant_type=QuantType.per_1x32,
        doweight_stage1=False,
        w1_scale=gate_up_weight_scale_shuffled,
        w2_scale=down_weight_scale_shuffled,
        a1_scale=None,
        a2_scale=None,
        hidden_pad=hidden_pad,
        intermediate_pad=intermediate_pad,
    )