paddle的nansum不支持empty的求和
w5688414 opened this issue · comments
bug描述 Describe the Bug
>>> import paddle
>>> x = paddle.to_tensor([])
x>>> x.nansum()
W0517 13:31:49.496955 61548 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.0, Driver API Version: 11.8, Runtime API Version: 11.8
W0517 13:31:49.522248 61548 gpu_resources.cc:164] device: 0, cuDNN Version: 8.6.
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/root/paddlejob/workspace/output/wugaosheng/wu391/lib/python3.9/site-packages/paddle/tensor/math.py", line 1762, in nansum
return sum(tmp_tensor, axis, dtype, keepdim, name)
File "/root/paddlejob/workspace/output/wugaosheng/wu391/lib/python3.9/site-packages/paddle/tensor/math.py", line 1534, in sum
return _C_ops.sum(x, axis, dtype, keepdim)
ValueError: (InvalidArgument) Tensor need be reduced must not empty.
[Hint: Expected x.numel() > 0, but received x.numel():0 <= 0:0.] (at ../paddle/phi/kernels/funcs/reduce_function.h:1055)
>>> x.is_empty()
Tensor(shape=[], dtype=bool, place=Place(cpu), stop_gradient=True,
True)
>>> import torch
>>> x1 = torch.Tensor([])
>>> x1.nansum()
tensor(0.)
其他补充信息 Additional Supplementary Information
无
你好，从语义上来说，empty Tensor的数值计算应该属于未定义行为，请问目前对这个计算，有具体的需求场景吗