search

pyro-ppl / pyro

Deep universal probabilistic programming with Python and PyTorch
http://pyro.ai
Apache License 2.0
8.51k stars 984 forks source link

Problem with gp.util.train() in bayesian optimization #2314

Open simonegarrappa opened 4 years ago

simonegarrappa commented 4 years ago

Hi, I am not able to reproduce the Bayesian Optimization snippet from https://pyro.ai/examples/bo.html

In particular, the code gets stuck running continuously without handling errors when executing the command

gp.util.train(gpmodel, optimizer)

the issue is present even specifying a very low number of steps.

My specs are: Mac OSX Mojave 10.14.6 Pytorch: 1.4.0 Pyro: 1.2.1

Thanks in advance. Best, Simone

fehiepsi commented 4 years ago

@simonegarrappa Did you resolve the issue? I can't replicate it in my system and we have a test for its execution in Travis. This might be an edge case due to the system I guess.

simonegarrappa commented 4 years ago

@fehiepsi seems like it is indeed a system related issue. I'm able to run it on google colab.