Hi,

To date, I code all of my deep learning models from scratch. I found a library made using standard published paper which will give me complete control over the parameters of the model.

So, I have a few questions. If I have a good theoretical understanding of how the model works, then should I:

1. Use a library where everything is already coded

2. Build from scratch (as this will give me more confidence while using the model)

I just want your view. I want to be a researcher/academician. What should be my approach?