Mergekit: "The Best of All Worlds"

The idea with mergekit is that you can essentially take two models of similar architecture with the same amount of hidden layers, then smash them together, into a new model.

Sometimes, it makes a monster, other times, it makes a masterpiece.

You don't need to know what you're doing to run a mergekit script...but to use it in a productive way...that takes a relative understanding of the models you're working with.

To get Mergekit installed, go to the github repository (opens in a new tab) and just follow the Installation header.

You don't have to install it though, you can use LazyMergekit instead: https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing#scrollTo=ik0V0dF55gfU (opens in a new tab)