Hi Misgana, Can you commit all the changes to your fork of carbon-ml, and share us the link. So that I can check whats happening.
Also, do I need to manually edit the machinelearner.xml file(s) to include > Stacking? Yes. No need to add hyper-parameters, since we can't predefine them. Thanks, Supun On Mon, Jun 27, 2016 at 4:33 AM, Misgana Negassi < [email protected]> wrote: > Hello Supun, > sorry to bother you again with the same problem but I am not making any > progress right now :( > I can't get the sample script running. > > I tried two approaches: > A) Using the server provided and applying patches. > Problem: I can not apply my patch. I suspect the problem is that my patch > is of a newer version. The server's carbon ml is 1.1.1. and mine is 1.1.2. > Can I somewhere change the version number when I build my project? > > B) Building carbon and then product-ml and then unpacking > modules/distribution/target/wso2ml-2.0.0-SNAPSHOT.zip > I can start the server with the right components except for one exception: > > log4j:ERROR Could not instantiate appender named "CARBON_MEMORY". > log4j:ERROR Could not instantiate class > [org.wso2.carbon.logging.service.appender.CarbonMemoryAppender]. > java.lang.ClassNotFoundException: > org.wso2.carbon.logging.service.appender.CarbonMemoryAppender cannot be > found by org.wso2.carbon.logging_4.4.3 > at > org.eclipse.osgi.internal.loader.BundleLoader.findClassInternal(BundleLoader.java:501) > > > However, when I try to run my example script there seems to be nothing > happening on the server side. No exceptions but also the debugger does not > stop when I expect it would. It seems something is crashing but the > exceptions are swallowed. Is there a way to increase the debug output? > > Also, do I need to manually edit the machinelearner.xml file(s) to include > Stacking? > > > On 21.06.2016 05:41, Supun Sethunga wrote: > > Hi Misgana, > > I have two issues: > 1. Issue: > > Exception in thread "pool-17-thread-7" java.lang.IllegalArgumentException: >> No enum constant org.wso2.carbon.ml.commons.con >> stants.MLConstants.SUPERVISED_ALGORITHM.STACKING >> at java.lang.Enum.valueOf(Enum.java:238) >> at org.wso2.carbon.ml.commons.constants.MLConstants$SUPERVISED_ >> ALGORITHM.valueOf(MLConstants.java:225) >> at org.wso2.carbon.ml.core.spark.algorithms.SupervisedSparkMode >> lBuilder.build(SupervisedSparkModelBuilder.java:158) > > Which approach did you try? putting the jar as a patch (patch9999)? or > build the carbon-ml and product ml respectively? If it was patching method, > did you build org.wso2.carbon.ml.commons as well? > > Caused by: org.wso2.carbon.ml.database.exceptions.DatabaseHandlerException: >> An error occurred while inserting hyper parameter to the database: Value >> too long for column "VALUE VARCHAR(50)": "'rO0ABXNyABNqYXZhLnV0aWwuQXJy >> YXlMaXN0eIHSHZnHYZ0DAAFJAARzaXpleHAAAAACdwQAAAACc3IAEWphdmEu >> dXRpbC5IYXNoTWFwBQfawcMWYNEDAAJGAApsb2F... (372)"; SQL statement: >> INSERT INTO ML_HYPER_PARAMETER(ANALYSIS_ID, ALGORITHM_NAME, NAME, VALUE, >> LAST_MODIFIED_TIME) VALUES(?,?,?,?, CURRENT_TIMESTAMP()) [90005-140] >> The Logic is: The hyperparameters of base-algorithms are serialized and >> passed in the product-ml hyperparamters script file.This is a serialized >> string of List of Maps, where each map contains hyperparameters of each >> base-algorithm and will be deserialized in buildStackingModel() in >> SupervisedSparkModelBuilder and fed to Stacking train(). >> Current idea to solve it is: >> 1. Increase column capacity more than VALUE VARCHAR(50) >> 2. Change serialization logic >> What do you think? > > > I believe we don't need to pass a serialized a map and put the string in > DB. Can we store the hyperparamerets as follows, in the ML_HYPER_PARAMETER > table? > > *ANALYSIS_ID* *ALGORITHM_NAME* *NAME* *VALUE* *LAST_MODIFIED_TIME* > 123 Meta_Algorithm_1 Learning Rate 0.001 <some_timestamp> > 123 Meta_Algorithm_1 Num_Trees 10 <some_timestamp> > 123 Meta_Algorithm_1 Max_Depth 0.001 <some_timestamp> > 123 Meta_Algorithm_1 Seed 4567 <some_timestamp> > 123 Meta_Algorithm_2 Learning Rate 0.001 <some_timestamp> > 123 Meta_Algorithm_2 Iterations 100 <some_timestamp> > 123 Meta_Algorithm_2 Seed 789 <some_timestamp> > … … … … … > 123 Base_Algorithm Learning Rate 0.001 <some_timestamp> > 123 Base_Algorithm Iterations 100 <some_timestamp> > 123 Base_Algorithm Seed 6325 <some_timestamp> > > Do you see any complications/issues in doing so? > > > > Thanks, > Supun > > On Mon, Jun 20, 2016 at 6:58 PM, Misgana Negassi < > [email protected]> wrote: > >> Hi Supun, >> >> I have two issues: >> 1. Issue: >> >> Exception in thread "pool-17-thread-7" >> java.lang.IllegalArgumentException: No enum constant >> org.wso2.carbon.ml.commons.constants.MLConstants.SUPERVISED_ALGORITHM.STACKING >> at java.lang.Enum.valueOf(Enum.java:238) >> at >> org.wso2.carbon.ml.commons.constants.MLConstants$SUPERVISED_ALGORITHM.valueOf(MLConstants.java:225) >> at >> org.wso2.carbon.ml.core.spark.algorithms.SupervisedSparkModelBuilder.build(SupervisedSparkModelBuilder.java:158) >> >> >> It doesn't recognize STACKING although I have built carbonml as you have >> described. >> >> 2. Issue >> >> >> Caused by: >> org.wso2.carbon.ml.database.exceptions.DatabaseHandlerException: An error >> occurred while inserting hyper parameter to the database: Value too long >> for column "VALUE VARCHAR(50)": >> "'rO0ABXNyABNqYXZhLnV0aWwuQXJyYXlMaXN0eIHSHZnHYZ0DAAFJAARzaXpleHAAAAACdwQAAAACc3IAEWphdmEudXRpbC5IYXNoTWFwBQfawcMWYNEDAAJGAApsb2F... >> (372)"; SQL statement: >> INSERT INTO ML_HYPER_PARAMETER(ANALYSIS_ID, ALGORITHM_NAME, NAME, VALUE, >> LAST_MODIFIED_TIME) VALUES(?,?,?,?, CURRENT_TIMESTAMP()) [90005-140] >> >> The Logic is: The hyperparameters of base-algorithms are serialized and >> passed in the product-ml hyperparamters script file.This is a serialized >> string of List of Maps, where each map contains hyperparameters of each >> base-algorithm and will be deserialized in buildStackingModel() in >> SupervisedSparkModelBuilder and fed to Stacking train(). >> >> Current idea to solve it is: >> 1. Increase column capacity more than VALUE VARCHAR(50) >> 2. Change serialization logic >> >> What do you think? >> >> Regards, >> Misgana >> >> >> >> >> Can you specify why what was the error? Is there any error/stack trace? >>> >> >> > > > -- > *Supun Sethunga* > Senior Software Engineer > WSO2, Inc. > <http://wso2.com/>http://wso2.com/ > lean | enterprise | middleware > Mobile : +94 716546324 > Blog: <http://supunsetunga.blogspot.com>http://supunsetunga.blogspot.com > > > -- *Supun Sethunga* Senior Software Engineer WSO2, Inc. http://wso2.com/ lean | enterprise | middleware Mobile : +94 716546324 Blog: http://supunsetunga.blogspot.com
_______________________________________________ Dev mailing list [email protected] http://wso2.org/cgi-bin/mailman/listinfo/dev
