[developers] What is the most memory consumed by a parse reranking process on the logon tree?

[Bill McNeill (UW)]

 What was your memory high water mark during test (as opposed to training)?


On Mon, Feb 23, 2009 at 9:07 PM, Francis Bond <fcbond at gmail.com> wrote:

> G'Day,
>
> 2009/2/24 Bill McNeill (UW) <billmcn at u.washington.edu>:
> > I'm trying to run the parse reranking scripts for English as described
> here
> > (http://wiki.delph-in.net/moin/LogonModeling).  Various jobs keep dying
> with
> > out of memory errors.  What is the largest amount of RAM one of these
> > processes is going to consume?
>
> On a similar process for Japanese we would occasionally run out of
> memory on an 8GB machine (when doing the EM step).  On a 32GB machine
> we had no problems.
> The feature caching I would expect to need up to 4GB at least.
>
> --
> Francis Bond <http://www2.nict.go.jp/x/x161/en/member/bond/>
> NICT Language Infrastructure Group
>



-- 
Bill McNeill
http://staff.washington.edu/billmcn/index.shtml
Sent from: Seattle Washington United States.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20090224/6ba7ff88/attachment.html>