Excellent. That's exactly what I was looking for. Thanks, Lane On Tue, May 24, 2011 at 2:33 PM, Philipp Koehn <[email protected]> wrote:
> Hi, > > if you set it to 0, then no threshold pruning takes place. > > The relevant lines are in HypothesisStackNormal.cpp: > if ( m_bestScore + m_beamWidth > m_worstScore ) > m_worstScore = m_bestScore + m_beamWidth; > > where m_beamWidth are the log of the value you specify. > If you set it to 0, then it is -inf. > > -phi > > On Mon, May 23, 2011 at 11:48 PM, Lane Schwartz <[email protected]> > wrote: > > So would it be meaningful to set this value to zero? > > In my current experiment, I want to ensure that no beam threshold pruning > > takes place. > > Thanks, > > Lane > > > > On Mon, May 23, 2011 at 5:46 PM, Philipp Koehn <[email protected]> > wrote: > >> > >> Hi, > >> > >> you should make it smaller (e.g., 0.000000001f). > >> > >> Hypotheses whose score is worse than the best in a stack > >> by this factor are dropped. > >> > >> -phi > >> > >> On Mon, May 23, 2011 at 7:50 PM, Lane Schwartz <[email protected]> > wrote: > >> > I'm trying to understand how to set the beam-threshold parameter. The > >> > default value is 0.00001f. > >> > > >> > If I want this parameter to prune less, should I make the value large > >> > (for > >> > example, 0.1) or make it smaller (for example 0.000000001f)? > >> > > >> > Thanks, > >> > Lane > >> > > >> > _______________________________________________ > >> > Moses-support mailing list > >> > [email protected] > >> > http://mailman.mit.edu/mailman/listinfo/moses-support > >> > > >> > > > > > > > > > -- > > When a place gets crowded enough to require ID's, social collapse is not > > far away. It is time to go elsewhere. The best thing about space travel > > is that it made it possible to go elsewhere. > > -- R.A. Heinlein, "Time Enough For Love" > > > -- When a place gets crowded enough to require ID's, social collapse is not far away. It is time to go elsewhere. The best thing about space travel is that it made it possible to go elsewhere. -- R.A. Heinlein, "Time Enough For Love"
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
