How to work with OutputDistances object?

[hsm207]

In simsimd 3.7.7, this code:

import simsimd
import numpy as np

X = np.array([[1, 2, 3]], dtype=np.float32)
Y = np.array([[4, 5, 6]], dtype=np.float32)

simsimd.cosine(X, Y)

returns

array([0.02940369], dtype=float32)

but in 3.9.0, it now returns:

<simsimd.OutputDistances at 0xffff6257ec70>

how do I get the result like in 3.7.7?

[ashvardanian]

Hi, @hsm207! Have you tried wrapping via np.array?

[ashvardanian]

I had to change this behavior to output a generic buffer instead of NumPy array, because the latter supports fewer platforms than SimSIMD, so we can’t always rely on NumPy types. The provided output, however, must be safely convertible to any tensor type in NumPy, PyTorch, TensorFlow, or any other numeric library.

I’m sorry for the inconvenience it may have caused, I didn’t realize many depended on the exact type. I can probably added to the README. Feel free to a documentation patch if you have a minute 🤗

[maiqingqiang]

Hi, @hsm207! Have you tried wrapping via np.array?

How to wrap it with np.array? Is there a reference example?

[ashvardanian]

Have you tried simply passing the output into the np.array function?

[ashvardanian]

Commit c09d895 explicitly describes conversion rules. Hope this helps 🤗