submission 566241

"""
First baseline submission: directly call AITER gemm_a4w4.
This matches the reference implementation exactly.
"""
import torch
from task import input_t, output_t

import aiter
from aiter import dtypes
from aiter.ops.triton.quant import dynamic_mxfp4_quant
from aiter.utility.fp4_utils import e8m0_shuffle


def custom_kernel(data: input_t) -> output_t:
    A, B, B_q, B_shuffle, B_scale_sh = data
    A = A.contiguous()

    # 1) Quant A to MXFP4 with shuffled scales
    A_fp4, A_scale_e8m0 = dynamic_mxfp4_quant(A)
    A_q = A_fp4.view(dtypes.fp4x2)
    A_scale_sh = e8m0_shuffle(A_scale_e8m0).view(dtypes.fp8_e8m0)

    # 2) GEMM a4w4
    out = aiter.gemm_a4w4(
        A_q,
        B_shuffle,
        A_scale_sh,
        B_scale_sh,
        dtype=dtypes.bf16,
        bpreshuffle=True,
    )
    return out