假设我想计算两个张量之间的元素商。如果其中一个张量包含NaN,则生成的商也将包含NaN,这一点我是理解的。但是为什么在整个操作中梯度变得不存在呢?如何为非NaN条目保留渐变?
例如:
>>> x = torch.tensor([1.0, np.NaN])
>>> y = torch.tensor([2.0, 3.0])
>>> z = torch.div(y, x)
>>> z
tensor([2., nan])
>>> z.backward()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/usr/local/lib/python3.6/dist-packages/torch/tensor.py", line 107, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph)
File "/usr/local/lib/python3.6/dist-packages/torch/autograd/__init__.py", line 93, in backward
allow_unreachable=True) # allow_unreachable flag
RuntimeError: element 0 of tensors does not require grad and does not have a grad_fn
发布于 2019-06-26 05:09:50
这只是一个小调整:
import torch
import numpy as np
x = torch.tensor([1.0, np.NaN], requires_grad=True)
y = torch.tensor([2.0, 3.0])
z = torch.div(y, x)
z #tensor([2., nan], grad_fn=<DivBackward0>)
换句话说,你需要说你需要梯度计算,否则张量将没有梯度函数。
https://stackoverflow.com/questions/56761865
复制相似问题