Hi,

I'm trying to run Moses through its experiment management system across 
a cluster. I have a step (LM:news:tokenize) where the default time limit 
the queue is giving it is not enough, so I need to pass an argument to 
qsub to specify a new time limit.

If I specify the flag in qsub-settings(*) under [GENERAL] or [LM] or 
[LM:news] it has no effect. However, if I specify it under the heading 
[LM:news:tokenize] it causes the step LM:news:tokenize to be submitted 
to the queue twice.

(*) I'm using qsub-settings rather than qsub-hours because the cluster 
is running Torque rather than the Sun Grid Engine and the command-line 
argument used when specifying qsub-hours is not available.


The relevant extract from the config file when it's submitting twice is:

----
[LM:news]
raw-corpus = $wmt10-data/training/news.$output-extension.shuffled

[LM:news:tokenize]
qsub-settings = "-l walltime=24:00:00"
----

whereas this causes it to be submitted to the queue once, but without 
the time limit increase:

----
[LM:news]
raw-corpus = $wmt10-data/training/news.$output-extension.shuffled

qsub-settings = "-l walltime=24:00:00"
----

Why would this be, and does anyone know how I might fix it?

Thanks,
Suzy

-- 
Suzy Howlett
http://www.showlett.id.au/
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to