gistfile1.txt

In [1]: import faiss

In [2]: import numpy as np

In [4]: d = 32

In [5]: m = 4

In [6]: c = 6

In [7]: pq = faiss.ProductQuantizer(d, m, c)

In [8]: opq = faiss.OPQMatrix(d, m)

In [10]: opq.pq = pq

In [12]: opq.verbose = True

In [14]: x = np.random.rand(100, d).astype('float32')

In [15]: opq.train(x)
OPQMatrix::train: training an OPQ rotation matrix for M=4 from 100 vectors in 32D -> 32D
  OPQMatrix::train: making random 32*32 rotation
Training PQ slice 0/4
WARNING clustering 100 points to 64 centroids: please provide at least 2496 training points
Clustering 100 points in 8D to 64 clusters, redo 1 times, 40 iterations
  Preprocessing in 0.00 s
  Iteration 39 (0.49 s, search 0.35 s): objective=8.0993 imbalance=1.318 nsplit=0
Training PQ slice 1/4
WARNING clustering 100 points to 64 centroids: please provide at least 2496 training points
Clustering 100 points in 8D to 64 clusters, redo 1 times, 40 iterations
  Preprocessing in 0.00 s
  Iteration 39 (0.04 s, search 0.03 s): objective=8.07424 imbalance=1.318 nsplit=0
  
  ....
Training PQ slice 2/4
WARNING clustering 100 points to 64 centroids: please provide at least 2496 training points
Clustering 100 points in 8D to 64 clusters, redo 1 times, 4 iterations
  Using 64 centroids provided as input (not frozen)
  Preprocessing in 0.00 s
  Iteration 3 (0.00 s, search 0.00 s): objective=4.14809 imbalance=1.306 nsplit=0
Training PQ slice 3/4
WARNING clustering 100 points to 64 centroids: please provide at least 2496 training points
Clustering 100 points in 8D to 64 clusters, redo 1 times, 4 iterations
  Using 64 centroids provided as input (not frozen)
  Preprocessing in 0.00 s
  Iteration 3 (0.00 s, search 0.00 s): objective=4.65961 imbalance=1.254 nsplit=0
    encode / decode
    Iteration 49 (4 PQ iterations):1.730 s, obj=0.178715
    X * recons