Great work! I want to get the rules, but it seems cost too much GPU memory.
tingxueronghua opened this issue · comments
I think your work is excellent and takes lots of efforts! I try to get the rules in my server, however, the cuda is out of memory in line 182 of rule_mining...
Is it possible and convenient for you to provide the rules? And I want to inquire that is it possible to use less CUDA memory to reproduce your work?
To reduce GPU memory, you can either reduce batch size, or reduce the minimum support for gminer.
Line 182 is the transaction matrix. You can also keep it in cpu and do next operations in cpu also, but it might take more time.
I will try to release the rules when I have the time.
Thank you very much! I will try adapting the code as soon as possible! Looking forward to the release of rules!