确定性问题描述:[功能问题-EZ1001][torch.real]torch.real算子存在 功能问题-EZ1001 通路:torchair golden版本:20261103 golden链接:https://gitcode.com/opencann/operator-artifacts/blob/master/Real/golden文件/real.py cann版本:20261103 ttk版本:20261103 用例链接:https://gitcode.com/sleeveen/operator_test_case/tree/main/LongTail/Cases/Functional_Accuracy/torchair/torch.real 用例名:aclnnReal_L0_003 错误码:EZ1001 复现命令:python3 -m ttk e2e -i 失败用例集_torch.real_torchair_功能失败-EZ1001.csv --dl=1 --plugin /home/CI/golden/torch.real/ --device-whitelist=0 -c -d --aclgraph --xpu-perf -t aclnnReal_L0_003 问题现象:
【用例 aclnnReal_L0_003】 2026-11-09 01:24:16,417 [WARNING] [2244469 aclnnReal_L0_003 OnEagerProfiling] [torch_profiler.py:215]: NPU profiler has no new kernel_details.csv for aclnnReal_L0_003 in /home/CI/ops-test-kit-master_1/msprof/e2e/aclnnReal_L0_003; historical traces are not reused. Check profiler export logs and directory permissions. 2026-11-09 01:24:24,635 [ERROR] [2244469 aclnnReal_L0_003 OnGraphAclgraph] [graph_runner.py:540]: Graph aclgraph execution failed: copy_d2d_baseformat_opapi:../torch_npu/csrc/aten/ops/op_api/CopyKernelOpApi.cpp:125 NPU function error: call aclnnInplaceCopy failed, error code is 161002 [ERROR] 2026-11-09-01:24:16 (PID:2244469, Device:0, RankID:-1) ERR00100 PTA call acl api failed. [ERROR] 2026-11-09-01:24:16 (PID:2244469, Device:0, RankID:-1) ERR00100 PTA call acl api failed.
CANNBot实际失败分析: 1) 根因与类别
dst.copy_(src)
aclnnInplaceCopy
op_broadcast
AclNN_Parameter_Error(EZ1001): Tensor self not implemented for DT_COMPLEX32, should be in dtype support list [DT_INT8...DT_COMPLEX64,DT_COMPLEX128...]
acl_graph.py:1535 process_input: dst_tensors[0].copy_(src_tensors[0])
Failed to get TorchAir NPU backend: cannot import name 'op_broadcast'
EAGER PASS
迁移来源:https://gitcode.com/sleeveen/operator_test_case/issues/233
附件下载链接:
确定性问题描述:[功能问题-EZ1001][torch.real]torch.real算子存在 功能问题-EZ1001
通路:torchair
golden版本:20261103
golden链接:https://gitcode.com/opencann/operator-artifacts/blob/master/Real/golden文件/real.py
cann版本:20261103
ttk版本:20261103
用例链接:https://gitcode.com/sleeveen/operator_test_case/tree/main/LongTail/Cases/Functional_Accuracy/torchair/torch.real
用例名:aclnnReal_L0_003
错误码:EZ1001
复现命令:python3 -m ttk e2e -i 失败用例集_torch.real_torchair_功能失败-EZ1001.csv --dl=1 --plugin /home/CI/golden/torch.real/ --device-whitelist=0 -c -d --aclgraph --xpu-perf -t aclnnReal_L0_003
问题现象:
CANNBot实际失败分析:
1) 根因与类别
dst.copy_(src)拷贝 complex32 输入,底层aclnnInplaceCopy不支持 DT_COMPLEX32(支持列表只有 complex64/128),属 NPU适配实现缺陷(dtype支持不全),非golden/用例问题(eager对golden已PASS)。op_broadcast导入失败,属 环境/版本配套问题,与本体例失败无关但需修复。2) 关键证据
AclNN_Parameter_Error(EZ1001): Tensor self not implemented for DT_COMPLEX32, should be in dtype support list [DT_INT8...DT_COMPLEX64,DT_COMPLEX128...]acl_graph.py:1535 process_input: dst_tensors[0].copy_(src_tensors[0])Failed to get TorchAir NPU backend: cannot import name 'op_broadcast'EAGER PASS(golden dtype float16)3) 定位建议
开发责任人:叶钊
PL:叶钊
迁移来源:https://gitcode.com/sleeveen/operator_test_case/issues/233