ThunderSVM fit for very long time

[AnsonHuangTW]

Hi everyone :)

These days, I found sometimes it going like a deadlock when I fit the svc model. I tried sklearn, libsvm, libsvm with gpu with same data, and it can fit in 8 seconds. But I can't find what's happening in thunderSVM (both thunderSVM gpu, cpu). Can anyone give me some suggestions to solve this problem?

Cuda: 10.2 Python: 3.9.7 Code&Data

Really appreciate any help at all!