Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.
Atlas 训练系列产品卷积反向1d，fmap=kernel的case运行失败，需要修复
Atlas 训练系列产品
构造卷积反向1d，fmap=kernel的case运行
Atlas 训练系列产品卷积反向1d，fmap=kernel的case运行成功
详细信息见附件 Atlas 训练系列产品卷积反向1d，fmap=kernel的case运行 [INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.632 [unsqueeze.cpp:318][NNOP][SplitStringAndPrint][43445] OpName:[aclnnConvolutionBackward_3_UnsqueezeNd] Entering function params: x: aclTensor(view_shape = [3, 5, 1, 6], view_strides = [30, 6, 6, 1], view_offset = 0, view_format = NCHW, storage_shape = [3, 5, 1, 6], storage_format = NCHW, ori_shape = [3, 5, 1, 6], ori_format = NCHW, data_type = DT_FLOAT, device_ptr = 0x12e83c000800), dim: [2]. [INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.642 [unsqueeze.cpp:260][NNOP][OpDfxGuard][43445] OpName:[aclnnConvolutionBackward_3_UnsqueezeNd] Entering function end. 0 [INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.652 [unsqueeze.cpp:56][NNOP][UnsqueezeNd][43445] OpName:[aclnnConvolutionBackward_3_UnsqueezeNd] Leaving function UnsqueezeNd [INFO] ASCENDCL(43355,python3.12):2026-03-20-16:03:05.479.661 [device.cpp:138]43445 aclrtGetDeviceImpl: start to execute aclrtGetDevice [DEBUG] ASCENDCL(43355,python3.12):2026-03-20-16:03:05.479.669 [device.cpp:145]43445 aclrtGetDeviceImpl: successfully execute aclrtGetDevice, get device id is 5. [INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.679 [platform.cpp:410][NNOP][GetPlatformInfo][43445] OpName:[aclnnConvolutionBackward_3] GetCurrentPlatformInfo, get current device is: 5, tPlatformInfoList.size(): 6 [ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.688 [aclnn_convolution_backward.cpp:2452][NNOP][CalculateConv3DBp][43445] errno[161002] OpName:[aclnnConvolutionBackward_3] not implemented for 1001 [INFO] GE(43355,python3.12):2026-03-20-16:03:05.479.774 [error_manager.cc:451]43445 ReportErrMessage:report error_message, error_code:EZ1001, work_stream_id:4335543445, error_mode:1. [ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.804 [aclnn_convolution_backward.cpp:2506][NNOP][CalculateConv2DBp][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check ret == ACLNN_SUCCESS failed. [ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.815 [aclnn_convolution_backward.cpp:1427][NNOP][CalculateConv1DBackward][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check status == ACLNN_SUCCESS failed. [ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.826 [aclnn_convolution_backward.cpp:2543][NNOP][CalculateConvolutionBackward][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check ret == ACLNN_SUCCESS failed. [ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.835 [aclnn_convolution_backward.cpp:2712][NNOP][aclnnConvolutionBackwardGetWorkspaceSize][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check ret == ACLNN_SUCCESS failed.
/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)
Atlas 训练系列产品卷积反向1d，fmap=kernel的case运行失败，需要修复
Environment / 环境信息 (Mandatory / 必填)
Atlas 训练系列产品
Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)
构造卷积反向1d，fmap=kernel的case运行
Describe the expected behavior / 预期结果 (Mandatory / 必填)
Atlas 训练系列产品卷积反向1d，fmap=kernel的case运行成功
Related log / screenshot / 日志 / 截图 (Mandatory / 必填)
详细信息见附件
Atlas 训练系列产品卷积反向1d，fmap=kernel的case运行
[INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.632 [unsqueeze.cpp:318][NNOP][SplitStringAndPrint][43445] OpName:[aclnnConvolutionBackward_3_UnsqueezeNd] Entering function params: x: aclTensor(view_shape = [3, 5, 1, 6], view_strides = [30, 6, 6, 1], view_offset = 0, view_format = NCHW, storage_shape = [3, 5, 1, 6], storage_format = NCHW, ori_shape = [3, 5, 1, 6], ori_format = NCHW, data_type = DT_FLOAT, device_ptr = 0x12e83c000800), dim: [2].
[INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.642 [unsqueeze.cpp:260][NNOP][OpDfxGuard][43445] OpName:[aclnnConvolutionBackward_3_UnsqueezeNd] Entering function end. 0
[INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.652 [unsqueeze.cpp:56][NNOP][UnsqueezeNd][43445] OpName:[aclnnConvolutionBackward_3_UnsqueezeNd] Leaving function UnsqueezeNd
[INFO] ASCENDCL(43355,python3.12):2026-03-20-16:03:05.479.661 [device.cpp:138]43445 aclrtGetDeviceImpl: start to execute aclrtGetDevice
[DEBUG] ASCENDCL(43355,python3.12):2026-03-20-16:03:05.479.669 [device.cpp:145]43445 aclrtGetDeviceImpl: successfully execute aclrtGetDevice, get device id is 5.
[INFO] OP(43355,python3.12):2026-03-20-16:03:05.479.679 [platform.cpp:410][NNOP][GetPlatformInfo][43445] OpName:[aclnnConvolutionBackward_3] GetCurrentPlatformInfo, get current device is: 5, tPlatformInfoList.size(): 6
[ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.688 [aclnn_convolution_backward.cpp:2452][NNOP][CalculateConv3DBp][43445] errno[161002] OpName:[aclnnConvolutionBackward_3] not implemented for 1001
[INFO] GE(43355,python3.12):2026-03-20-16:03:05.479.774 [error_manager.cc:451]43445 ReportErrMessage:report error_message, error_code:EZ1001, work_stream_id:4335543445, error_mode:1.
[ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.804 [aclnn_convolution_backward.cpp:2506][NNOP][CalculateConv2DBp][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check ret == ACLNN_SUCCESS failed.
[ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.815 [aclnn_convolution_backward.cpp:1427][NNOP][CalculateConv1DBackward][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check status == ACLNN_SUCCESS failed.
[ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.826 [aclnn_convolution_backward.cpp:2543][NNOP][CalculateConvolutionBackward][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check ret == ACLNN_SUCCESS failed.
[ERROR] OP(43355,python3.12):2026-03-20-16:03:05.479.835 [aclnn_convolution_backward.cpp:2712][NNOP][aclnnConvolutionBackwardGetWorkspaceSize][43445] errno[561000] OpName:[aclnnConvolutionBackward_3] check ret == ACLNN_SUCCESS failed.
Special notes for this issue/备注 (Optional / 选填)