Next release for Stan?

[Daniel Lee]

Anyone oppose to us releasing a new version of the Stan library soon? Bob fixed a couple bugs that's on develop now.


Daniel

[Allen B. Riddell]

Sounds good. If you fix a date in advance I can try to have PyStan ready
by then.

On 02/18, Daniel Lee wrote:
> Anyone oppose to us releasing a new version of the Stan library soon? Bob fixed a couple bugs that's on develop now.
>
>
> Daniel
>

> --
> 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/d/optout.

[Bob Carpenter]

And we have the new CVODE solver. Or are we not
going to expose that right away?

- Bob

> On Feb 18, 2016, at 5:00 PM, Daniel Lee <bea...@alum.mit.edu> wrote:
>
> Anyone oppose to us releasing a new version of the Stan library soon? Bob fixed a couple bugs that's on develop now.
>
>
> Daniel
>

[Michael Betancourt]

I thought CVODE was already in by 2.8.2 — or do you mean
exposing it through the interfaces properly?

[Sebastian Weber]

Nope, CVODE is not out there which is why relasing some time soon would be great. Just to bring things wrt to dependecies on the radar which would be good to address before release (I suppose):

lib / Stan / CRAN version
boost / 1.58.0 / 1.60.0
eigen / 3.2.4 / 3.2.7

As far as I can tell this is not a problem, but an alignment is certainly better, i.e. boost 1.60.0 works with Stan, but you get a ton of deprecated messages.

Sebastian

On Friday, February 19, 2016 at 9:54:25 AM UTC+1, Michael Betancourt wrote:
> I thought CVODE was already in by 2.8.2 — or do you mean
> exposing it through the interfaces properly?
>
> On Feb 19, 2016, at 3:17 AM, Bob Carpenter>

> > And we have the new CVODE solver. Or are we not
> > going to expose that right away?
> >
> > - Bob
> >
> >> On Feb 18, 2016, at 5:00 PM, Daniel Lee > >>

[Michael Betancourt]

No, CVODE has absolutely been related in math/stan/cmdstan.
Getting CVODE installed in RStan and PyStan has not yet been
done, but that is an interface issue, not a Stan issue.

[Bob Carpenter]

I hadn't even realized CVODE had gone is. The
interfaces being out of sync with respect to which
language features they support makes writing the
manual tricky. We're going to have to start
qualifying features by release/platform.

- Bob

[Krzysztof Sakrejda]

On Friday, February 19, 2016 at 12:08:00 PM UTC-5, Bob Carpenter wrote:
> I hadn't even realized CVODE had gone is. The
> interfaces being out of sync with respect to which
> language features they support makes writing the
> manual tricky. We're going to have to start
> qualifying features by release/platform.
>
> - Bob

We could just put a big ugly feature matrix on the website for discrepancies
between interfaces. The language manual is complicated enough without
including interface/version-specific cruft.

Krzysztof

[Michael Betancourt]

Or interfaces expose all of the features of Stan before
working on their own specialized features...

[Ben Goodrich]

On Friday, February 19, 2016 at 12:22:46 PM UTC-5, Michael Betancourt wrote:

Or interfaces expose all of the features of Stan before
working on their own specialized features...

Or make it clear what needs to be implemented. Like Bob, I didn't know there was something that needed to be done with CVODE until after the Stan release for NIPS. At that time, the main thing I was aware of that needed to be done for rstan was to enable ADVI. There was a compiler error involving cvode.h when building against the develop branch, so I built against the master branch after 2.9.0, and it was fine. The compiler error with develop is still there

https://api.travis-ci.org/jobs/110429566/log.txt?deansi=true

so I'll have to do some more stuff with the paths in order for cvode_2.8.2/ to be searched.

Ben 
 

>>>>>> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

>>>>>> For more options, visit https://groups.google.com/d/optout.
>>>>>>
>>>>>
>>>>> --
>>>>> 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+unsubscribe@googlegroups.com.

>>>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>> --
>>> 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+unsubscribe@googlegroups.com.

>>> For more options, visit https://groups.google.com/d/optout.
>
> --
> 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+unsubscribe@googlegroups.com.

[Allen B. Riddell]

Couldn't we just infer some credible interval concerning the features
the interfaces are likely to support?