oneflow.amax perform differently between cpu and cuda when dim=1

[Redmept1on]

Summary

oneflow.softmax perform differently between cpu and cuda when dim=1

Code to reproduce bug

import oneflow as flow
import numpy as np

x1 = flow.tensor(np.array([[float('inf'), 0, -1, float('nan'), 5]], dtype=np.float32))
x1 = x1.cuda()
y1 = flow.amax(x1,dim=1)
print(y1.device,y1)

x1 = flow.tensor(np.array([[float('inf'), 0, -1, float('nan'), 5]], dtype=np.float32))
x1 = x1.cpu()
y2 = flow.amax(x1,dim=1)
print(y2.device,y2)

System Information

• What is your OneFlow installation :dockerhub
• OS:Linux
• OneFlow version (run python3 -m oneflow --doctor):0.9.1.dev20231024+cu117
• Python version: 3.8

[Redmept1on]

oneflow.amin has the same issue

import oneflow as flow
import numpy as np

x1 = flow.tensor(np.array([[float('inf'), 0, -1, float('nan'), 5]], dtype=np.float32))
x1 = x1.cuda()
y1 = flow.amin(x1,dim=1)
print(y1.device,y1)

x1 = flow.tensor(np.array([[float('inf'), 0, -1, float('nan'), 5]], dtype=np.float32))
x1 = x1.cpu()
y2 = flow.amin(x1,dim=1)
print(y2.device,y2)