defining random state or seed option parameters

[kahramankostas]

I need the option to assign random state or seed values to get stable results. I don't think there is such an option. Unfortunately, my attempts to fix the general seed that I have listed below did not solve the problem.

import random
random.seed(1)
from numpy.random import seed
seed(1)

What can be done about it? Do you have any advice? thanks in advance

[VHRanger]

Thanks.

Are you looking at node2vec specifically? I agree that all the algorithms should support it, however. For Node2vec we'd have to manually seed for the random walks, and once again on when training the network. I don't think it would be too much work to do.

[VHRanger]

Seems like the seeding has to be wrapped by numba jit code, see this issue: https://github.com/numba/numba/issues/6002

(Random walks are numba-jitted)

The long term fix will be in the CSRGraphs package as well

[VHRanger]

Having looked deeper into it, this is more complex than it seems - we use numba.prange in loops for threading, so setting a numba seed before the multithreaded loop doesn't guarantee replicable results.

It's seemingly unresolved on the numba side; see this issue:

https://github.com/numba/numba/issues/4452