DriverlessAI Custom Scorers

Packages

Scorer	Package	Version	Download
Database	TAR	2.30	Download
Hive UDF	TAR	2.0	Download
REST Server	TAR	1.0	Download

DB Scorer

Requirements

1. Driverless AI Mojo

-- copy the mojo file in the models/ directory

2. Driverless AI License

-- provided via the partnership portal

-- copy the license to the license.sig file in the license/ directory

-- run export DRIVERLESS_AI_LICENSE_FILE=/path/to/license.sig

3. Java JDK 1.8

-- execute “java -version” to verify the available JVM on your platform has JDK 1.8 installed. The output of the command should indicate JDK 1.8

4. Properties File

-- create one or use the default properties file under properties/ directory

5. JAR File

-- DAI-Mojo-DB jar file in scorer/ directory

Execution

java -Derrors=true -Dverbose=true -Dthreads=2 -Dcapacity=4000 -Dpropertiesfilename=DAIMojoRunner_DB.properties -jar DAI-Mojo-DB-2.30.jar

Result

-- The DB scorer stores the score/prediction for each row from the given dataset in a file called results.csv in the main directory.

-- Create a results.csv file in the main directory (if it doesn't exist already). It is important for the execution command to have a results.csv file in the main dir to save the data processed by the scorer as the code will not create the file itself if it doesn't exist.

Note:

• Set -Dthreads=200 and -Xms10g for faster processing
• You must have a DAI license to score, add the license in one of the standard ways, as a parameter, as a environment variable -Dai.h2o.mojos.runtime.license.file=/Driverless-AI/license/license.sig

Hive Scorer

Requirements

1. Driverless AI Mojo

-- copy the mojo file in the models/ directory

2. Driverless AI License

-- provided via the partnership portal

-- copy the license to the license.sig file in the license/ directory

-- run export DRIVERLESS_AI_LICENSE_FILE=/path/to/license.sig

3. Java JDK 1.8

-- execute “java -version” to verify the available JVM on your platform has JDK 1.8 installed. The output of the command should indicate JDK 1.8

4. H2O Runtime Drivers

-- runtime drivers in lib/ directory

5. JAR File

-- DAI-Mojo-Hive jar file in scorer/ directory

Execution

In hive environment run:

hive> add jar /path/to/mojo2-runtime.jar;
Added [/home/munish/DAI-Mojo-Hive/lib/mojo2-runtime.jar] to class path
Added resources: [/home/munish/DAI-Mojo-Hive/lib/mojo2-runtime.jar]

hive> add jar /path/to/DAI-Mojo-Hive-.jar;
Added [/home/munish/DAI-Mojo-Hive/scorer/DAI-Mojo-Hive-.jar] to class path
Added resources: [/home/munish/DAI-Mojo-Hive/scorer/DAI-Mojo-Hive-.jar]

hive> list jar;
/home/munish/DAI-Mojo-Hive/lib/mojo2-runtime.jar
/home/munish/DAI-Mojo-Hive/scorer/DAI-Mojo-Hive-.jar

hive> create temporary function daiPredict as 'ai.h2o.mojos.hive.daiPredict';
OK
Time taken: 0.75 seconds

hive> set DRIVERLESS_AI_LICENSE_FILE=/path/to/license.sig;

select id, daiPredict("DRIVERLESS_AI_MODELNAME=/path/to/mojo/pipeline.mojo verbose",loan_amnt, term, int_rate, installment, emp_length, home_ownership, annual_inc, verification_status, addr_state, dti, delinq_2yrs, inq_last_6mths, pub_rec, revol_bal, revol_util, total_acc) from lcdata where addr_state='CA' and loan_amnt>34000 limit 1;
calling load
returned load
1068159 0.8994684115021625 0.10053158849783754
Time taken: 1.389 seconds, Fetched: 1 row(s)

hive> describe function extended daiPredict;
Call a Driverless AI Mojo for scoring prediction
Synonyms: daipredict

Rest Scorer

Requirements

1. Driverless AI Mojo

-- copy the mojo file in the models/ directory

2. Driverless AI License

-- provided via the partnership portal

-- copy the license to the license.sig file in the license/ directory

-- run export DRIVERLESS_AI_LICENSE_FILE=/path/to/license.sig

3. Java JDK 1.8

-- execute “java -version” to verify the available JVM on your platform has JDK 1.8 installed. The output of the command should indicate JDK 1.8

5. JAR File

-- DAI-Mojo2-RestServer jar file in scorer/ directory

Execution

java -DModelDirectory=models/pipeline.mojo -jar scorer/DAI-Mojo2-RestServer-1.0.jar

Results

curl "http://127.0.0.1:8080/model?name=pipeline.mojo&verbose=true&row=5000,36months,10.65,162.87,10,RENT,24000,VERIFIED-income,AZ,27.65,0,1,0,13648,83.7,0"