This is an unofficial implementation of the Recurrent Geometric Network as described in the paper End-to-End Differentiable Learning of Protein Structure. It is still a work in progress. I haven't tested the model on the ProteinNet dataset that the paper uses, so I cannot confirm that the model successfully reproduces the reported results.
At the current stage this repository is primarily useful as a guide. It does contain a correct, working model in the rgn notebook. In the next few weeks, I expect to reproduce the results of the paper, cleanup my code, and generally just make this repository more useful.