sum算子bug

[Difers]

在调试Beta算子时发现了sum算子的一个bug，
gpu环境下出错，cpu正常：测试代码如下

import paddle
import numpy as np
paddle.set_device("gpu")
array = np.array([], dtype=np.float32)
x = paddle.to_tensor(np.reshape(array, [0, 2]), dtype='float32')
x.sum(-1)

应该是reduce_ function.h下Run函数某一步出错，但下面的维度变化我没看懂QAQ

Originally posted by @Difers in #49922 (comment)

[wuyefeilin]

x为零维tensor, sum算子目前不支持该类型计算。

[Difers]

x为零维tensor, sum算子目前不支持该类型计算。

cpu环境下是支持的，想要修复的话是处理至和CPU环境下一致的结果，还是直接在两端都加上维度检查呢？

[wuyefeilin]

我看在cpu下面x.sum()结果时错的，x.sum(-1)符合逻辑结构。要修复的话还是得在cpu和gpu下都进行修复，使其符合逻辑

[luotao1]

@Difers 可以尝试修复一下～

[paddle-bot]

Since you haven't replied for more than a year, we have closed this issue/pr.
If the problem is not solved or there is a follow-up one, please reopen it at any time and we will continue to follow up.
由于您超过一年未回复，我们将关闭这个issue/pr。
若问题未解决或有后续问题，请随时重新打开，我们会继续跟进。