Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.
卷积正向WieghtUbTrans模板在k轴尾块Ci不满足C0对齐时需要配置loopRpSize补0，补齐大小为Ci补 * kh * kw，若kh*kw较大，会超出loopRpSize的支持范围：[0-255]，三数值溢出不符合预期，导致部分内存未初始化，偶现精度失败nan。
Ascend950
不定期多次执行，偶现nan精度失败，与环境状态有关，参考规格如下： fmap：[15, 21, 257, 7] weight：[255, 21, 21, 3] pads：[13,13,13,13] strides：[6,6] dilations：[10,10] 没有bias，groups=1 输入输出dtype为bfloat16，format为NCHW
期望精度稳定通过
/assign
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)
卷积正向WieghtUbTrans模板在k轴尾块Ci不满足C0对齐时需要配置loopRpSize补0，补齐大小为Ci补 * kh * kw，若kh*kw较大，会超出loopRpSize的支持范围：[0-255]，三数值溢出不符合预期，导致部分内存未初始化，偶现精度失败nan。
Environment / 环境信息 (Mandatory / 必填)
Ascend950
Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)
不定期多次执行，偶现nan精度失败，与环境状态有关，参考规格如下：
fmap：[15, 21, 257, 7]
weight：[255, 21, 21, 3]
pads：[13,13,13,13]
strides：[6,6]
dilations：[10,10]
没有bias，groups=1
输入输出dtype为bfloat16，format为NCHW
Describe the expected behavior / 预期结果 (Mandatory / 必填)
期望精度稳定通过
Related log / screenshot / 日志 / 截图 (Mandatory / 必填)
Special notes for this issue/备注 (Optional / 选填)