New loss function : bayesian loss

[deebuls]

https://aleximmer.github.io/Laplace/

With the package u can convert any pytorch model to bayesian model and get uncertainty.

Please check it @SathwikSai10 .

There is a blog to explain the paper. https://towardsdatascience.com/go-deep-but-also-go-bayesian-ab25efa6f7b