Setting H2O FLOW directory path in Sparkling Water

Sometimes you may want to back up H2O FLOW files to some source code repo or to a backup location. For that reason you may want to change the default FLOW directory.

In H2O flag -flow_dir is used to set the local folder for FLOW files.

Note: You can always specify any H2O property by using system properties on Spark driver/executors.

So to change H2O FLOW directory to save you can append to your command line with the Sparkling Water commandline:

--conf spark.driver.extraJavaOptions="-Dai.h2o.flow_dir=/your/backup/location"

 

Thats it, thanks.

Using RESTful API to get POJO and MOJO models in H2O

 

CURL API for Listing Models:

http://<hostname>:<port>/3/Models/

CURL API for Listing specific POJO Model:

http://<hostname>:<port>/3/Models/model_name

List Specific MOJO Model:

http://<hostname>:<port>/3/Models/glm_model/mojo

Here is an example:

curl -X GET "http://localhost:54323/3/Models"
curl -X GET "http://localhost:54323/3/Models/deeplearning_model" >> NAME_IT

curl -X GET "http://localhost:54323/3/Models/deeplearning_model" >> dl_model.java
curl -X GET "http://localhost:54323/3/Models/glm_model/mojo" > myglm_mojo.zip

Thats it, enjoy!!

Starter script for rsparkling (H2O on Spark with R)

The rsparkling R package is an extension package for sparklyr that creates an R front-end for the Sparkling WaterSpark package from H2O. This provides an interface to H2O’s high performance, distributed machine learning algorithms on Spark, using R. Visit github project: https://github.com/h2oai/rsparkling

You must have the following package installed in your R environment:

You must have Sparkling Water latest package download and unzipped locally:

I am using the following package in my environment:

  • Spark 2.1
  • Sparkling Water 2.1.8
  • sparklyr 0.4.4
  • rsparkling 0.2.0

Now here is rspakrling script to create the cluster locally:

options(rsparkling.sparklingwater.location="/tmp/sparkling-water-assembly_2.11-2.1.8-all.jar")
Sys.setenv(SPARK_HOME="/usr/hdp/current/spark2-client/")
library(sparklyr)
library(rsparkling)
config <- spark_config()
config$spark.executor.cores <- 4
config$spark.executor.memory <- "4G"
sc <- spark_connect(master = "local", config = config, version = '2.1.0')
print(sc)
h2o_context(sc, strict_version_check = FALSE)
h2o_flow(sc, strict_version_check = FALSE)
spark_disconnect(sc)

Now here is the rsparkling script to create Spark cluster with Yarn:

options(rsparkling.sparklingwater.location="/tmp/sparkling-water-assembly_2.11-2.1.8-all.jar")
Sys.setenv(SPARK_HOME="/usr/hdp/current/spark2-client/")
library(sparklyr)
library(rsparkling)
config <- spark_config()
config$spark.executor.cores <- 4
config$spark.executor.memory <- "4G"
config$spark.executor.instances = 2
sc <- spark_connect(master = "yarn-client", config = config, version = '2.1.0')
print(sc)
h2o_context(sc, strict_version_check = FALSE)
h2o_flow(sc, strict_version_check = FALSE)
spark_disconnect(sc)

Thats it, Enjoy!!

Binomial classification example in Scala and GBM with H2O

Here is a sample for binomial classification problem using H2O GBM algorithm using Credit Card data set in Scala language.

The following sample is for multinomial classification problem. This sample is created using Spark 2.1.0 with Sparkling Water 2.1.4.

import org.apache.spark.h2o._
import water.support.SparkContextSupport.addFiles
import org.apache.spark.SparkFiles
import java.io.File
import water.support.{H2OFrameSupport, SparkContextSupport, ModelMetricsSupport}
import water.Key
import _root_.hex.glm.GLMModel
import _root_.hex.ModelMetricsBinomial


val hc = H2OContext.getOrCreate(sc)
import hc._
import hc.implicits._

addFiles(sc, "/Users/avkashchauhan/learn/deepwater/credit_card_clients.csv")
val creditCardData = new H2OFrame(new File(SparkFiles.get("credit_card_clients.csv")))

val ratios = Array[Double](0.8)
val keys = Array[String]("train.hex", "valid.hex")
val frs = H2OFrameSupport.split(creditCardData, keys, ratios)
val (train, valid) = (frs(0), frs(1))

def buildGLMModel(train: Frame, valid: Frame, response: String)
 (implicit h2oContext: H2OContext): GLMModel = {
 import _root_.hex.glm.GLMModel.GLMParameters.Family
 import _root_.hex.glm.GLM
 import _root_.hex.glm.GLMModel.GLMParameters
 val glmParams = new GLMParameters(Family.binomial)
 glmParams._train = train
 glmParams._valid = valid
 glmParams._response_column = response
 glmParams._alpha = Array[Double](0.5)
 val glm = new GLM(glmParams, Key.make("glmModel.hex"))
 glm.trainModel().get()
 //val glmModel = glm.trainModel().get()
}

val glmModel = buildGLMModel(train, valid, 'default_payment_next_month)(hc)

// Collect model metrics and evaluate model quality
val trainMetrics = ModelMetricsSupport.modelMetrics[ModelMetricsBinomial](glmModel, train)
val validMetrics = ModelMetricsSupport.modelMetrics[ModelMetricsBinomial](glmModel, valid)
println(trainMetrics.rmse)
println(validMetrics.rmse)
println(trainMetrics.mse)
println(validMetrics.mse)
println(trainMetrics.r2)
println(validMetrics.r2)
println(trainMetrics.auc)
println(validMetrics.auc)

// Preduction
addFiles(sc, "/Users/avkashchauhan/learn/deepwater/credit_card_predict.csv")
val creditPredictData = new H2OFrame(new File(SparkFiles.get("credit_card_predict.csv")))

val predictionFrame = glmModel.score(creditPredictData)
var predictonResults = asRDD[DoubleHolder](predictionFrame).collect.map(_.result.getOrElse(Double.NaN))

Thats it, enjoy!!

Binomial classification example in Scala and GLM with H2O

Here is a sample for binomial classification problem using H2O GLM algorithm using Credit Card data set in Scala language.

The following sample is for multinomial classification problem. This sample is created using Spark 2.1.0 with Sparkling Water 2.1.4.

import org.apache.spark.h2o._
import water.support.SparkContextSupport.addFiles
import org.apache.spark.SparkFiles
import java.io.File
import water.support.{H2OFrameSupport, SparkContextSupport, ModelMetricsSupport}
import water.Key
import _root_.hex.glm.GLMModel
import _root_.hex.ModelMetricsBinomial


val hc = H2OContext.getOrCreate(sc)
import hc._
import hc.implicits._

addFiles(sc, "/Users/avkashchauhan/learn/deepwater/credit_card_clients.csv")
val creditCardData = new H2OFrame(new File(SparkFiles.get("credit_card_clients.csv")))

val ratios = Array[Double](0.8)
val keys = Array[String]("train.hex", "valid.hex")
val frs = H2OFrameSupport.split(creditCardData, keys, ratios)
val (train, valid) = (frs(0), frs(1))

def buildGLMModel(train: Frame, valid: Frame, response: String)
 (implicit h2oContext: H2OContext): GLMModel = {
 import _root_.hex.glm.GLMModel.GLMParameters.Family
 import _root_.hex.glm.GLM
 import _root_.hex.glm.GLMModel.GLMParameters
 val glmParams = new GLMParameters(Family.binomial)
 glmParams._train = train
 glmParams._valid = valid
 glmParams._response_column = response
 glmParams._alpha = Array[Double](0.5)
 val glm = new GLM(glmParams, Key.make("glmModel.hex"))
 glm.trainModel().get()
 //val glmModel = glm.trainModel().get()
}

val glmModel = buildGLMModel(train, valid, 'default_payment_next_month)(hc)

// Collect model metrics and evaluate model quality
val trainMetrics = ModelMetricsSupport.modelMetrics[ModelMetricsBinomial](glmModel, train)
val validMetrics = ModelMetricsSupport.modelMetrics[ModelMetricsBinomial](glmModel, valid)
println(trainMetrics.rmse)
println(validMetrics.rmse)
println(trainMetrics.mse)
println(validMetrics.mse)
println(trainMetrics.r2)
println(validMetrics.r2)
println(trainMetrics.auc)
println(validMetrics.auc)

// Prediction
addFiles(sc, "/Users/avkashchauhan/learn/deepwater/credit_card_predict.csv")
val creditPredictData = new H2OFrame(new File(SparkFiles.get("credit_card_predict.csv")))

val predictionFrame = glmModel.score(creditPredictData)
var predictonResults = asRDD[DoubleHolder](predictionFrame).collect.map(_.result.getOrElse(Double.NaN))

 

Thats it, enjoy!!

Using one Hot Encoding and Stratified Sampling in Sparkling Water & H2O

Here is an example Scala code which shows how to use new Sparkling Water package to setup OneHot encoding and Stratified Sampling:

import org.apache.spark.h2o._
import water.Key
import water.etl.prims.advmath.AdvMath.StratifiedSplit
import water.etl.prims.mungers.Mungers.OneHotEncoder
import water.etl.prims.operators.Operators.Eq
import water.fvec.Frame
import water.fvec.H2OFrame
import java.net.URI

// Use the following line if you decided to setup external SW Kluster
// val conf = new H2OConf(sc).setExternalClusterMode().useManualClusterStart().setCloudName("test")
//val hc = H2OContext.getOrCreate(sc, conf)

// OR use the following for basic configuration
val hc = H2OContext.getOrCreate(sc)
val fr = new H2OFrame(new java.io.File("/Users/avkashchauhan/src/github.com/h2oai/h2o-3/smalldata/airlines/AirlinesTest.csv.zip"))
val frOH = OneHotEncoder(fr, "Origin")
fr.add(frOH)   // Combine the pivoted result to the original frame
val trainTestCol = StratifiedSplit(fr,"IsDepDelayed",0.2,123);
val idx = Eq(trainTestCol,"train")
val train = fr.deepSlice(idx,null)   

// get subset of the Frame according to True/False of boolean 1 column Frame "idx"
val idx2 = Eq(trainTestCol,"test")
val test = fr.deepSlice(idx2,null)
println(train.toString(0L,10))
Thats it, enjoy!!

Creating, Adding and managing H2O frame in Scala

Creating a new H2O Frame:

To create a new frame in H2O you will call as below:

val df = new Frame()

Adding a frame to another H2O Frame:

To add an H2O from to another H2O frame you will do the following:
frame1.add(frame2)
When h2oDataFrame.add() method is called, it mutates the calling frame. It doesn’t create a new Frame and the Frame keeps the same Key.  Its the same object in memory.
What happens is that frame1 now depends on frame2.  Frame “frame1” has the new columns but they are actually the data from “frame2”. Looking into this operation, it looks like data has been duplicated because there are 2 keys in the DKV, but actually there has been no memory copy at all.  If you delete “frame2” you will run into an error , because the Frame “frame1” now depends on “frame2”.
In general managing memory in H2O DKV, there is no automated way of deleting old Frames during your program execution, you just need to manually call Frame.delete() on the Frames you no longer need.

Difference of using val vs var in Scala with new frame:

While looking from Scala point of view val dataframeNew = new Frame() doesn’t stop you from changing the dataframeNew frame with dataframeNew.add,  this does however stop you from reassigning dataframeNew to a different instance of a Frame.
Note: If you had var dataframeNew = new Frame(), then this df can be set to a completely different Frame. The reason for this difference is mainly because how Scala treats the val vs var in variable assignment.
Thats it, enjoy