How to set random seed.

[JKYUPSYCHE]

Hi. I want to set random seed for reproducibility of sampling.
I used 'random.seed( )', but I don't think it's working.
How can I set random seed in HSSM?

[AlexanderFengler]

Hi @JKYUPSYCHE,

thank you for trying HSSM.

This may help.
Let us know if it doesn't resolve the issue.

Best,
Alex

[JKYUPSYCHE]

Thank you Alex, and sorry for the late reply.
I tried, but It's not working yet.
sampling code is here.

rng = np.random.default_rng(42)

model_test = hssm.HSSM(
model="ddm",
...
)

model_test.sample(idata_kwargs=dict(
log_likelihood=True
), random_seed=rng)

This code doesn't cause an error, but the output value is slightly different.

[digicosmos86]

Hi @JKYUPSYCHE,

Per this page, and since HSSM is a wrapper around bambi, you can use the random_seed argument directly when you call model.sample(). This argument is simply an int or a list of ints, depending on how many cores you are using. In other words, you need to set a random seed for each core that you are using when sampling.

Let us know if this works.

Thanks!
Paul

[JKYUPSYCHE]

Hi Paul,
Now It's working! Thank you for you answer.