Hi,
we have trouble fit in model in 10GB for breast. Maybe our base image is having some size problem, but we are reluctant to change any code at this stage to fit in the 10GB basically we will have to train half of the genes off line and half of the genes on line. we will need to split and concatenate the files. this kind of change could easily make result random.
any advice for that?
Yuanfang
Created by Yuanfang Guan 小黄药师 yuanfang.guan ok. never mind, my teammate just said he would split and train offline half and online half.
but still i feel then it becomes an engineering problem, which is certainly a great advantage for my own team , but half of the teams won't be able to do this step correctly to get a non-random result on the new data that hasn't be leaderboarded before.