Implementation of Memorizing Transformers (ICLR 2022), attention net augmented with indexing and retrieval of memories using approximate nearest neighbors, in Pytorch
Do you know of any solutions to efficiently solve the "hard reset" problem in FAISS? I know that one could use IndexFlatL2 but that's not really efficient.
Hello and thanks for this implementation!
Do you know of any solutions to efficiently solve the "hard reset" problem in FAISS? I know that one could use IndexFlatL2 but that's not really efficient.
Thank you!