search

Xtra-Computing / thundersvm

ThunderSVM: A Fast SVM Library on GPUs and CPUs
Apache License 2.0
1.57k stars 218 forks source link

Debug Assertion Failed on SVM.predict with Probability = 1 #246

Closed bmd2007 closed 3 years ago

bmd2007 commented 3 years ago

I'm testing out my install of SVM predict on windows using cuda, but when I set probability to true, the SVC fit function crashes with an assertion error

Here is the code I'm currently using

import thundersvm s = thundersvm.SVC(probability=1) #I've tried with probability=True and probability=1, both error import numpy as np d = np.random.random((100,10)) c = np.random.random((100)) s.fit(d,c)

Error Message:

Debug Assertion Failed! Program: ...packages\thundersvm\thundersvm\build\bin\debug\thundersvm.dll File: C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.28.29333\include\vector Line: 1511

Expression: vector subscript out of range

bmd2007 commented 3 years ago

I should note that the code runs find when I don't use the probability argument (only binary classification)

bmd2007 commented 3 years ago

Stupid error. It couldn't handle floating point classes. Sorry.