NPE in ALS

[illes]

NPE in ALS

[baltrunasl]

Maria,
could you look at this, or I start?
I can send you the data if you want and command on which it fails.

Thanks!

[illes]

The computation seems to be stuck in an infinite loop:

14/03/27 11:17:14 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:19 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:24 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:29 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:34 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:39 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 442.79MB, average 442.79MB
14/03/27 11:17:44 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 442.79MB, average 442.79MB
...

[dlogothetis]

I think it's a problem with how the job was launched. The computation
specified should be ml.grafos.okapi.cf.als.Als.InitUsersComputation. The
very first superstep must run this computation to initialize the vertices
properly. Apart from this you also need to specify the master compute class
to use.

More info on how to run it correctly here:
http://grafos.ml/okapi.html#collaborative-als

On Thu, Mar 27, 2014 at 12:18 PM, illes [email protected] wrote:

The computation seems to be stuck in an infinite loop:

14/03/27 11:17:14 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:19 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:24 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:29 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:34 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 444.16MB, average 444.16MB
14/03/27 11:17:39 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 442.79MB, average 442.79MB
14/03/27 11:17:44 INFO job.JobProgressTracker: Data from 1 workers - Compute superstep 0: 0 out of 85 vertices computed; 0 out of 1 partitions computed; min free memory on worker 1 - 442.79MB, average 442.79MB

Reply to this email directly or view it on GitHubhttps://github.com//issues/6#issuecomment-38791284
.

[illes]

My fault, I adapted the run configuration by simply changing the algo name and forgot to specify:

ml.grafos.okapi.cf.als.Als\$InitUsersComputation \
-mc ml.grafos.okapi.cf.als.Als\$MasterCompute \

as suggested in the documentation: http://grafos.ml/okapi.html#collaborative-als