Refactoring HarDNet for readability

[joeltio]

Hi, I would like to work on making the HarDNet code more readable so that others can understand and adapt it more easily.

I have started working on some simple things to improve the code such as error handling for unsupported architectures or clearing trailing whitespace.

Currently I have these things in mind:

• Updating get_links to be more concise and also generalise to more than 1024 layers (because of this
• More comments to explain what is going on and where the implementation was mentioned in the paper
• Update loading of model to support cpu-based models (this part needs a map_location=torch.device("cpu") for cpu-based models to work)
• General readability of other parts of the code

Please let me know if there's any part you might also want me to look at or if there's something you would like to keep.

[PingoLH]

Hi, Thank you so much for your contribution and so sorry for the late reply. Please let me know if your change is ready to merge and I will merge it right away.