oneflow.gt/ge/greater/greater_equal perform differently between cpu and cuda.

[Redmept1on]

Summary

oneflow.gt/ge/greater/greater_equal perform differently between cpu and cuda. maybe from the same root cause.

Code to reproduce bug

Take greater_equal as an example

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()
x2 = flow.tensor(np.array([float('nan'), 0, -1, 1, 5], dtype=np.float32))
x2 = x2.cuda()
y1 = flow.greater_equal(x1,x2)
print(y1)

x1 = flow.tensor(np.array([float('inf'), 0, -1, float('nan'), 5], dtype=np.float32))
x1=x1.cpu()
x2 = flow.tensor(np.array([float('nan'), 0, -1, 1, 5], dtype=np.float32))
x2 = x2.cpu()
y2 = flow.greater_equal(x1,x2)
print(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