SearchParametersIVF does not seem to work with IndexHNSWFlat

[eigenvectorBazuz]

I use IndexHNSWFlat to search a small dataset (1024 entries) of dimension 64 and it works fine. But when I try to use the SearchParametersIVF option to restrict some of the searches I get the following error:

`---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
/tmp/ipykernel_3079933/2324586828.py in
----> 1 M_b1 = calc_faiss_mask_causal_blocks(query_data, key_data, k=65, blocks=1, mode='graph')
2 evaluate_mask(query_data, key_data, M_b1, k=65)
3 # print(np.sum(Mf, axis=1))
4 # print(Mf[0,:])

~/cs-gpt/notebooks/lsh_aux.py in calc_faiss_mask_causal_blocks(Q, K, k, blocks, mode)
378 D, I = index.search(Q[query_block_range,:], k=k-1, params=faiss.SearchParametersIVF(sel=id_selector, nprobe=3))
379 elif mode == 'graph':
--> 380 D, I = index.search(Q[query_block_range,:], k=k-1, params=faiss.SearchParametersIVF(sel=id_selector))
381
382 # print(I)

~/.local/lib/python3.9/site-packages/faiss/class_wrappers.py in replacement_search(self, x, k, params, D, I)
341 assert I.shape == (n, k)
342
--> 343 self.search_c(n, swig_ptr(x), k, swig_ptr(D), swig_ptr(I), params)
344 return D, I
345

~/.local/lib/python3.9/site-packages/faiss/swigfaiss_avx2.py in search(self, n, x, k, distances, labels, params)
6424 def search(self, n, x, k, distances, labels, params=None):
6425 r""" entry point for search"""
-> 6426 return _swigfaiss_avx2.IndexHNSW_search(self, n, x, k, distances, labels, params)
6427
6428 def reconstruct(self, key, recons):

RuntimeError: Error in virtual void faiss::IndexHNSW::search(faiss::idx_t, const float*, faiss::idx_t, float*, faiss::idx_t*, const faiss::SearchParameters*) const at /project/faiss/faiss/IndexHNSW.cpp:297: Error: 'params' failed: params type invalid`

[mdouze]

As the name indicates, you should use SearchParametersHNSW, that also has an IDSelector field.