precomputed_knn error

Nobody has claimed this yet.

api documentation here https://umap-learn.readthedocs.io/en/latest/api.html tells :

" If you wish to use k-nearest neighbors data calculated by another package then provide a tuple of the form (knn_indices, knn_dists). The contents of the tuple should be two numpy arrays of shape (N, n_neighbors) where N is the number of items in the input data. The first array should be the integer indices of the nearest neighbors, and the second array should be the corresponding distances. The nearest neighbor of each item should be itself, e.g. the nearest neighbor of item 0 should be 0, the nearest neighbor of item 1 is 1 and so on. Please note that you will not be able to transform new data in this case. "

If you provide such an input to umap api it raises this error during fit :

File /opt/conda/lib/python3.10/site-packages/umap/umap_.py:2288, in UMAP.fit(self, X, y)
2286 self.knn_indices = self.precomputed_knn[0]
2287 self.knn_dists = self.precomputed_knn[1]
-> 2288 self.knn_search_index = self.precomputed_knn[2]
2290 self._validate_parameters()
2292 if self.verbose:

IndexError: tuple index out of range

if this input is provided (knn_indices, knn_dists,None) it raises :

File /opt/conda/lib/python3.10/site-packages/umap/umap_.py:2290, in UMAP.fit(self, X, y)
2287 self.knn_dists = self.precomputed_knn[1]
2288 self.knn_search_index = self.precomputed_knn[2]
-> 2290 self._validate_parameters()
2292 if self.verbose:
2293 print(str(self))

File /opt/conda/lib/python3.10/site-packages/umap/umap_.py:1930, in UMAP._validate_parameters(self)
1925 raise ValueError(
1926 "precomputed_knn[0] and precomputed_knn[1]"
1927 " must be numpy arrays of the same size."
1928 )
1929 if not isinstance(self.knn_search_index, NNDescent):
-> 1930 raise ValueError(
1931 "precomputed_knn[2] (knn_search_index)"
1932 " must be an NNDescent object."
1933 )
1934 if self.knn_dists.shape[1] < self.n_neighbors:
1935 warn(
1936 "precomputed_knn has a lower number of neighbors than "
1937 "n_neighbors parameter. precomputed_knn will be ignored"
1938 " and the k-nn will be computed normally."
1939 )

ValueError: precomputed_knn[2] (knn_search_index) must be an NNDescent object.

Is the documentation right ? Can we really provide precomputed_knn from other packages ?

Start with UMAP.fit and _validate_parameters in umap/umap_.py, then compare their handling of the documented precomputed_knn tuple with the API documentation. Reproduce the two errors using two- and three-element tuples; done means the documented input format and fit-time validation agree.

Written by the indexing model from the issue text.