Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[Breaking][jvm-packages] remove timeoutRequestWorkers parameter (#7839)

Browse Source
This commit is contained in:
Bobby Wang
2022-05-13 16:26:25 +08:00
committed by GitHub
parent 11d65fcb21
commit 9fa7ed1743
6 changed files with 7 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoost.scala
View File

@@ -75,7 +75,6 @@ private[scala] case class XGBoostExecutionParams(
missing: Float,
allowNonZeroForMissing: Boolean,
trackerConf: TrackerConf,
timeoutRequestWorkers: Long,
checkpointParam: Option[ExternalCheckpointParams],
xgbInputParams: XGBoostExecutionInputParams,
earlyStoppingParams: XGBoostExecutionEarlyStoppingParams,
@@ -201,12 +200,6 @@ private[this] class XGBoostExecutionParamsFactory(rawParams: Map[String, Any], s
case _ => throw new IllegalArgumentException("parameter \"tracker_conf\" must be an " +
"instance of TrackerConf.")
}
val timeoutRequestWorkers: Long = overridedParams.get("timeout_request_workers") match {
case None => 0L
case Some(interval: Long) => interval
case _ => throw new IllegalArgumentException("parameter \"timeout_request_workers\" must be" +
" an instance of Long.")
}
val checkpointParam =
ExternalCheckpointParams.extractParams(overridedParams)
@@ -227,7 +220,6 @@ private[this] class XGBoostExecutionParamsFactory(rawParams: Map[String, Any], s
val xgbExecParam = XGBoostExecutionParams(nWorkers, round, useExternalMemory, obj, eval,
missing, allowNonZeroForMissing, trackerConf,
timeoutRequestWorkers,
checkpointParam,
inputParams,
xgbExecEarlyStoppingParams,

2
jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostClassifier.scala
View File

@@ -66,8 +66,6 @@ class XGBoostClassifier (
def setMissing(value: Float): this.type = set(missing, value)
def setTimeoutRequestWorkers(value: Long): this.type = set(timeoutRequestWorkers, value)
def setCheckpointPath(value: String): this.type = set(checkpointPath, value)
def setCheckpointInterval(value: Int): this.type = set(checkpointInterval, value)

2
jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostRegressor.scala
View File

@@ -68,8 +68,6 @@ class XGBoostRegressor (
def setMissing(value: Float): this.type = set(missing, value)
def setTimeoutRequestWorkers(value: Long): this.type = set(timeoutRequestWorkers, value)
def setCheckpointPath(value: String): this.type = set(checkpointPath, value)
def setCheckpointInterval(value: Int): this.type = set(checkpointInterval, value)

11
jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/params/GeneralParams.scala
View File

@@ -112,15 +112,6 @@ private[spark] trait GeneralParams extends Params {
final def getAllowNonZeroForMissingValue: Boolean = $(allowNonZeroForMissing)
/**
* the maximum time to wait for the job requesting new workers. default: 30 minutes
*/
final val timeoutRequestWorkers = new LongParam(this, "timeoutRequestWorkers", "the maximum " +
"time to request new Workers if numCores are insufficient. The timeout will be disabled " +
"if this value is set smaller than or equal to 0.")
final def getTimeoutRequestWorkers: Long = $(timeoutRequestWorkers)
/**
* The hdfs folder to load and save checkpoint boosters. default: `empty_string`
*/
@@ -181,7 +172,7 @@ private[spark] trait GeneralParams extends Params {
setDefault(numRound -> 1, numWorkers -> 1, nthread -> 1,
useExternalMemory -> false, silent -> 0, verbosity -> 1,
customObj -> null, customEval -> null, missing -> Float.NaN,
trackerConf -> TrackerConf(), seed -> 0, timeoutRequestWorkers -> 30 * 60 * 1000L,
trackerConf -> TrackerConf(), seed -> 0,
checkpointPath -> "", checkpointInterval -> -1,
allowNonZeroForMissing -> false)
}

12
jvm-packages/xgboost4j-spark/src/test/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostGeneralSuite.scala
View File

@@ -165,18 +165,6 @@ class XGBoostGeneralSuite extends FunSuite with TmpFolderPerSuite with PerTest {
assert(x < 0.1)
}
test("training with spark parallelism checks disabled") {
val eval = new EvalError()
val training = buildDataFrame(Classification.train)
val testDM = new DMatrix(Classification.test.iterator)
val paramMap = Map("eta" -> "1", "max_depth" -> "6",
"objective" -> "binary:logistic", "timeout_request_workers" -> 0L,
"num_round" -> 5, "num_workers" -> numWorkers)
val model = new XGBoostClassifier(paramMap).fit(training)
val x = eval.eval(model._booster.predict(testDM, outPutMargin = true), testDM)
assert(x < 0.1)
}
test("repartitionForTrainingGroup with group data") {
// test different splits to cover the corner cases.
for (split <- 1 to 20) {