arg_refactor, not able to replicate with specifying seed

[Jiqiang Guo]

I might be making some mistakes.  But please try.  After I run the following in order, the outputs of two runs are different. 

  ./dogs data=dogs.data.R sample algorithm=hmc engine=static

  mv samples.csv samples.csv.old

  ./dogs data=dogs.data.R sample algorithm=hmc engine=static random seed=2874864247

  vimdiff samples.csv samples.csv.old

--

Jiqiang 

[Daniel Lee]

Is the first call missing a seed?

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Jiqiang Guo]

The first call does not specify the seed.  The second one uses the seed used in the first one.  So to test this, one cannot use the example I used. 

--

Jiqiang 

[Michael Betancourt]

I see it.  If you explicitly set a seed then it doesn't actually get passed to the RNG, instead 0 gets called.  The resultant behavior is that if you set a seed you get consistent results everytime, they're just consistent with a seed of 0 instead of the seed you specified (which is why I never noticed it).

Fixed, will push after I validate the tests.