There was indeed a problem with the mahout core job -- it did not include mahout-math in the lib directory. I've checked in a fix.
Could you point me to some sample data in arff format I could use as input to test the Random Forests code? Are the KDDTrain+.arff and KDDTrain+.info something you can share publicly? Thx, Drew On Thu, Oct 14, 2010 at 11:59 PM, deneche abdelhakim <[email protected]> wrote: > I had issues with random forests. I was testing the following example: > > https://cwiki.apache.org/MAHOUT/partial-implementation.html > > When I run the following command: > > $HADOOP_HOME/bin/hadoop jar > $MAHOUT_HOME/core/target/mahout-core-<VERSION>-job.jar > org.apache.mahout.df.tools.Describe -p testdata/KDDTrain+.arff -f > testdata/KDDTrain+.info -d N 3 C 2 N C 4 N C 8 N 2 C 19 N L > > I get a java.lang.NoClassDefFoundError: org/apache/mahout/math/Vector > > The same command works fine with mahout-0.3. > > Inside mahout-core-0.3.job you can find the Vector class inside > org/apache/mahout/mahout/, > but I can't seem to find it inside mahout-core-0.4-SNAPSHOT-job.jar > > am I missing something ? > > On Fri, Oct 15, 2010 at 3:38 AM, Drew Farris <[email protected]> wrote: >> Hi Deneche, Grant, >> >> There is an issue on jira related to this ( >> https://issues.apache.org/jira/browse/MAHOUT-505) >> >> The long and short of it is that nexus has problems with the way we were >> deploying artifacts that would prevent the jars for projects that produced >> job jars being deployed correctly. The job jar would be available when >> searching, but not the regular jar file. One way to work around this is to >> move from *.job to -job.jar >> >> This also allows us to use the maven assembly mechanism to build the job >> jars instead of using the ant build and maven build helper plug-in. There's >> nothing wrong with the approach pre-505, the post-505 approach just achieves >> the same goal with less configuration. >> >> As far as the Vector classes, the mahout-math jar is in the lib directory of >> the new job jars and thus available on the classpath when jobs are run using >> hadoop. >> >> Have you run into any issues using these new job jars? I've tested with the >> build-reuters.sh script and run bayes training haven't experienced any >> problems. >> >> Drew >> >> >> On Oct 14, 2010 12:57 PM, "Grant Ingersoll" <[email protected]> wrote: >> >
