Probabilistic-Programming-and-Bayesian-Methods-for-Hackers icon indicating copy to clipboard operation
Probabilistic-Programming-and-Bayesian-Methods-for-Hackers copied to clipboard

Published 20 hours ago • verified publisher icon CamDavidsonPilon

TFP versions | working with seeds

Open Pindar777 opened this issue 6 years ago • 0 comments

Hi @matthew-mcateer,

I tried to modify the code in order to get stable results. By doing so I came across this 'feature':

`intSeed = 12345 generate = tfd.Normal(0, 1)

1. try

evaluate(generate.sample(2, seed=intSeed)) # always: array([0.88424665, 0.07843047], dtype=float32)

2. try (intension to make code more readable

test = generate.sample(2, seed=intSeed) evaluate(test) # setting the seed does not work at all! ` IMHO using 'test' should not have this behavior. Do I miss something?

Thanks in advance

Pindar777 avatar Jan 13 '19 15:01 Pindar777