In this PR, I fine tune some hyper parameters of HRM & GRU4Rec & DIN. These models will hava a better perfromance with these parameters.
In GRU4Rec model, it is an effective way that setting epoch to a bigger number, such as 20. Small batch size number can also get a good effect. If you want to have better performance, you can try more epochs and smaller batch size.
In HRM model, I first set epoch to 10, but this change doesn't help. So I think the model may need more parameters. Then I set embedding to 64 and the model get a good performance.
In DIN model, it likes GRU4Rec model, small batch size number and more epochs are effective ways.
In this PR, I fine tune some hyper parameters of HRM & GRU4Rec & DIN. These models will hava a better perfromance with these parameters.
close #207 .