Hey Yash, hope you enjoyed the article. When implementing a paper i usually try to find as much info on how each module looks from a code perspective, like what layers to add and what activation to use as per the paper or similar papers. For example DCGAN paper is very descriptive on how the model architecture looks and what activation to use. You can also find some info from paperwithcode website for similar papers and try to adopt. Then try to tweak to the model to try to get the results as close as possible to that mentioned in the paper.

Hope this helps.