I’m trying to apply the concepts of the courses to a real test project, but when I start training (first epoch) my model accuracy is around 65%, and after 100 epoch is a little bit better 65.8%. My dataset is decent (600k examples) and the task is a binary classification, at the beginning I thought there were errors in my code, but if I fit a subset (10k examples) of the training data it learns at a decent rate. I’ve tried tuning the hyper parameters a bit, by increasing the learning rate and using a bigger and slightly deeper network, but it seems it doesn’t change much. Since a situation like this was not covered in the material, is it possible or there must be an error in my code? and if so what are the reason, are there possible solutions?