linfa-svm learning speed and memory allocation
clime opened this issue · comments
Hello,
I cannot get linfa-svm to produce any result as the operations take way too much memory and time:
Multi-class classifier full data:
[2023-07-11T08:18:38Z INFO be] Fit SVM classifier with #204242 training points
memory allocation of 333718356512 bytes failed
Aborted (core dumped)
Multi-class classifier limited data:
[2023-07-11T09:16:38Z INFO be] Fit SVM classifier with #10214 training points
^C
real 120m7.427s
user 119m52.661s
sys 0m2.573s
Regressor:
[2023-07-11T11:46:22Z INFO be] Fit SVM regressor with #10214 training points
(still running after 20 mins and will be probably running much, much longer)
I guess I could trim the input data set further but 10K input data points doesn't seem like way too many.
Is there any other way to speed things up?