Scale Up Bulk Similarity Calculations for Sparse Embeddings

ChunkDot support for sparse matrices

In my previous blog post, I introduced ChunkDot, a library that performs multi-threaded matrix multiplication and cosine similarity. ChunkDot is appropriate for calculating the K most similar items for a large number of items by chunking the item matrix representation (embeddings) and using Numba to accelerate the calculations.

Cosine Similarity for 1 Trillion Pairs of Vectors

I described how ChunkDot works under the hood and I showed some …

benchmarks blog bulk cosine-similarity data science embeddings library matrix matrix multiplication memory nlp numba representation scale support