Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.

Describe the current behavior / 问题描述 (Mandatory / 必填)

FlatQuant算子在mxfp4场景下，目前p2矩阵仅支持shape为(0,0)或(N,N),当前缺少对(0,N)和(N,0)等非法输入的拦截，需要增加拦截。

Environment / 环境信息 (Mandatory / 必填)

Ascend950

Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)

import torch
import torch_npu
K=16
M=64
N=64
x=torch.randn(K,M,N).half().npu()
p1=torch.randn(M,M).half().npu()
p2=torch.randn(0,N).half().npu()
print("p2.shpae: ",p2.shape)
clip_ratio=1.0
dst_dtype = torch_npu.float4_e2m1fn_x2
out, quant_scale = torch_npu.npu_kronecker_quant(x,p1,p2,clip_ratio,dst_dtype)

这里计算是没有拦截非法的p2

Describe the expected behavior / 预期结果 (Mandatory / 必填)

在p2是非法shape的时候，通过tiling拦截。

Related log / screenshot / 日志 / 截图 (Mandatory / 必填)

p2.shape异常的情况下没有拦截到

Special notes for this issue/备注 (Optional / 选填)