User Error 1001: omp_set_num_threads

[mattm401]

Running the windows build and during training/predicting on the latest push:

User Error 1001: omp_set_num_threads must be called with a positive integer 2018-05-14 14:35:02,451 INFO [default] #instances = 7136, #features = 4096 2018-05-14 14:35:03,938 INFO [default] #classes = 2 2018-05-14 14:36:02,830 INFO [default] working set size = 1024 2018-05-14 14:36:02,837 INFO [default] training start 2018-05-14 14:36:03,336 INFO [default] global iter = 0, total local iter = 512, diff = 2 2018-05-14 14:36:07,721 INFO [default] global iter = 48, total local iter = 16874, diff = 0.000993791 2018-05-14 14:36:07,728 INFO [default] rho = -0.00517282 2018-05-14 14:36:07,730 INFO [default] #sv = 7105 2018-05-14 14:36:10,845 INFO [default] #total unique sv = 7105 2018-05-14 14:36:10,912 INFO [default] training finished

Seem training program seems to be ignoring the OMP_NUM_THREADS set in my environmental variables, any idea how to resolve?

[mattm401]

clf = SVC(verbose=True, gamma=0.5, cost=100, n_cores=4)

Setting the n_cores took care of the message, but unclear if CPU or GPU cores?

[zeyiwen]

n_cores is to set the number of CPU cores. We'll inspect the warning.