You find the NNCLR model with code examples there: https://docs.lightly.ai/examples/nnclr.html
Please create any issues in the Lightly repository. This repository is not maintained anymore.
source/nn_memory_bank.py
implements a memory bank with nearest neigbours. In its forward pass, you input an output which updates the bank. Then the nearest neighbours to the output are returned.
nn_with_simsiam.py
is mostly the same file as from https://github.com/IgorSusmelj/simsiam-cifar10/blob/main/main.py. It is changed to replace the embeddings for out0 with their nearest neighbours from the memory bank.