ProductQuantizer compute_codes get wrong codes when nbits not 8

[jasstionzyf]

d = 10
n = 400000
cs = 5
np.random.seed(123)
x = np.random.random(size=(n, d)).astype('float32')
testInputs=np.random.random(size=(1, d)).astype('float32')
print(testInputs)
pq = faiss.ProductQuantizer(d, cs,6)
pq.verbose=True
pq.train(x)
codes=pq.compute_codes(testInputs)
#here expect 5 code range from 0-64, but get 4 and also code number not range 0-64
print(codes.shape)

[mdouze]

This is because the codes are packed into ceil(5 * 6 / 8) = 4 bytes.
To access the individual codes, use BitstringReader:

bs = faiss.BitstringReader(faiss.swig_ptr(codes[0]), codes.shape[1])
for i in range(cs): 
    print(bs.read(6))  # read 6 bits at a time

Admittedly, the BitstringReader API could be made more python friendly.

[jasstionzyf]

@mdouze thanks very much!