demo_merge_array_invertedlists.ipynb

Is it faster to successively query each small index or to query the merged large index（ When the CPU is fully used)?

@mdouze this doesn't work with multiple index. it only considers first index while searching hence the assertion is failing. Do you have a solution for it?

What do you mean with multiple index?

@mdouze The issue I'm encountering is if I give index_1 file, index_2 file , and index_3 file, if I serve them individually, the results are spread across them. After running the merging procedure I would expect the results to be the same as you also mentioned in your code (in cell 7 # make sure the results are the same) . However I see that tendentially, the search return items only included in the index_1 file (not in index_2 file and index_3 file) which eventually leads to assert error i.e. assert np.all(sI == I[i]). If I start with index_2 here (in cell 6 : faiss.extract_index_ivf(index).invlists ) then only it returns item from index_2. It seems like after merging, it just considers the first index file it encounters.