I’m attempting to train a model based on an old Kaggle competition that provides datasets on EEG results. I can’t really get the learner to train. When I run
learn .fit, I see the process using as much as 27GB of RAM. I’m thinking about experimenting with something like bcolz to further reduce memory usage, but I’m not certain this is what’s causing the training to stall/crash.
I suspect one issue might be that I need to make my dependent variables categorical somehow, but not quite sure how yet?
Here is my notebook so far:Basic EEG Model. Any help is appreciated! I:grinning: