This sounds a bit dubious to me. Mahout framework should provide a
reasonable default here.

My guess is that I discovered that -ow switch was not quite doing how it
was intended, so if you ran it multiple times, it might have been a
problem. it has been fixed in the latest round of fixes (0.8 trunk).
However, if it were a problem, your job would fail before it even launched
backend tasks, and i think it would be a different error message.

i would still try to look into task logs to see what's happening there. if
you dig that error out, i would be able to provide more info.


On Thu, Nov 15, 2012 at 4:50 AM, Abramov Pavel <[email protected]>wrote:

> I forgot to set --tempDir parameter.
>
>
> Issue Solved, thanks
>
>
> Pavel
>
>
> 15.11.12 15:43 пользователь "Abramov Pavel" <[email protected]>
> написал:
>
> >Hello!
> >
> >Trying to compute reduced rank matrix for recommendation system over
> >users (20*10^6) and their ratings (~150 000 items).
> >
> >How to avoid the exception with Q job? :
> >=======================
> >Exception in thread "main" java.io.IOException: Q job unsuccessful.
> >at org.apache.mahout.math.hadoop.stochasticsvd.QJob.run(QJob.java:230)
> >at
> >org.apache.mahout.math.hadoop.stochasticsvd.SSVDSolver.run(SSVDSolver.java
> >:377)
> >at
> >org.apache.mahout.math.hadoop.stochasticsvd.SSVDCli.run(SSVDCli.java:141)
> >=======================
> >
> >CLI parameters are:
> >=======================
> >mahout ssvd \
> >-i /rec/sparse/tfidf-vectors/ \
> >-o /rec/ssvd \
> >-k 100 \
> >--reduceTasks 100 \
> >-ow
> >=======================
> >Using Mahout 0.7 on Hadoop with ~50 nodes (400 Mappers/300 reducers).
> >
> >My Input for SSVD is seq2sparse output, and there are 200 "Key class:
> >class org.apache.hadoop.io.Text Value Class: class
> >org.apache.mahout.math.VectorWritable" sequence files. 8GB total.
> >
> >Many thanks in advance, any suggestion is highly appreciated. I Don't
> >know what to do, CF produces inaccurate results for my tasks, SVD is the
> >only hope ))
> >
> >Regards,
> >Pavel
> >
> >
> >
> >
>
>

Reply via email to