Save and Then Load

[lr1100]

Sorry if I missed in the docs or examples.

How would I go about saving off the pynndescent model to disk?
How would I load it from disk and run queries against it later (eval or operations)?

[Petkomat]

I simply use pickle from Python's standard library.

import pickle
import pynndescent

index = pynndescent.NNDescent(xs, ...)

with open("index.pkl", "wb") as f:
    pickle.dump(index, f)

with open("index.pkl", "rb") as f:
    reloaded_index = pickle.load(f)

[lr1100]

ah ok. maybe it is because we are using https://fdasrsf-python.readthedocs.io/en/latest/ for the metrics....

we are getting:
cannot pickle '_cffibackend.FFI' object.

thoughts? and thanks so much for replying.

[Petkomat]

Hi,

not everything can be pickled (see this for the same (?) error as above, and this for a slightly general explanation). However, unless you show the code (a minimal example that reproduces the error), it is hard to help.