Index pretransform support in search_preassigned (#3225)

Summary:
Pull Request resolved: https://github.com/facebookresearch/faiss/pull/3225

This diff fixes issue [#3113](https://github.com/facebookresearch/faiss/issues/3113), e.g. introduces support for index pretransform in `search_preassigned`.

Reviewed By: mdouze

Differential Revision: D53188584

fbshipit-source-id: 8189c0a59f957a2606391f22cf3fdc8874110a6e

contrib/ivf_tools.py

@@ -32,6 +32,11 @@ def search_preassigned(index_ivf, xq, k, list_nos, coarse_dis=None):
     Supports indexes with pretransforms (as opposed to the
     IndexIVF.search_preassigned, that cannot be applied with pretransform).
     """
+    if isinstance(index_ivf, faiss.IndexPreTransform):
+        assert index_ivf.chain.size() == 1, "chain must have only one component"
+        transform = faiss.downcast_VectorTransform(index_ivf.chain.at(0))
+        xq = transform.apply(xq)
+        index_ivf = faiss.downcast_index(index_ivf.index)
     n, d = xq.shape
     if isinstance(index_ivf, faiss.IndexBinaryIVF):
         d *= 8

tests/test_contrib.py

@@ -306,6 +306,26 @@ class TestRangeEval(unittest.TestCase):
 
 class TestPreassigned(unittest.TestCase):
 
+    def test_index_pretransformed(self):
+
+        ds = datasets.SyntheticDataset(128, 2000, 2000, 200)
+        xt = ds.get_train()
+        xq = ds.get_queries()
+        xb = ds.get_database()
+        index = faiss.index_factory(128, 'PCA64,IVF64,PQ4np')
+        index.train(xt)
+        index.add(xb)
+        index_downcasted = faiss.extract_index_ivf(index)
+        index_downcasted.nprobe = 10
+        xq_trans = index.chain.at(0).apply_py(xq)
+        D_ref, I_ref = index.search(xq, 4)
+
+        quantizer = index_downcasted.quantizer
+        Dq, Iq = quantizer.search(xq_trans, index_downcasted.nprobe)
+        D, I = ivf_tools.search_preassigned(index, xq, 4, Iq, Dq)
+        np.testing.assert_almost_equal(D_ref, D, decimal=4)
+        np.testing.assert_array_equal(I_ref, I)
+
     def test_float(self):
         ds = datasets.SyntheticDataset(128, 2000, 2000, 200)