Repository: spark Updated Branches: refs/heads/branch-2.0 dd70a115c -> 22b4072e7
[SPARK-16245][ML] model loading backward compatibility for ml.feature.PCA ## What changes were proposed in this pull request? model loading backward compatibility for ml.feature.PCA. ## How was this patch tested? existing ut and manual test for loading models saved by Spark 1.6. Author: Yanbo Liang <yblia...@gmail.com> Closes #13937 from yanboliang/spark-16245. (cherry picked from commit 0df5ce1bc1387a58b33cd185008f4022bd3dcc69) Signed-off-by: Xiangrui Meng <m...@databricks.com> Project: http://git-wip-us.apache.org/repos/asf/spark/repo Commit: http://git-wip-us.apache.org/repos/asf/spark/commit/22b4072e Tree: http://git-wip-us.apache.org/repos/asf/spark/tree/22b4072e Diff: http://git-wip-us.apache.org/repos/asf/spark/diff/22b4072e Branch: refs/heads/branch-2.0 Commit: 22b4072e704f9a68a605e9a4cebf54d2122fe448 Parents: dd70a11 Author: Yanbo Liang <yblia...@gmail.com> Authored: Tue Jun 28 19:53:07 2016 -0700 Committer: Xiangrui Meng <m...@databricks.com> Committed: Tue Jun 28 19:53:16 2016 -0700 ---------------------------------------------------------------------- .../scala/org/apache/spark/ml/feature/PCA.scala | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/spark/blob/22b4072e/mllib/src/main/scala/org/apache/spark/ml/feature/PCA.scala ---------------------------------------------------------------------- diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/PCA.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/PCA.scala index 72167b5..ef8b085 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/feature/PCA.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/feature/PCA.scala @@ -206,24 +206,22 @@ object PCAModel extends MLReadable[PCAModel] { override def load(path: String): PCAModel = { val metadata = DefaultParamsReader.loadMetadata(path, sc, className) - // explainedVariance field is not present in Spark <= 1.6 - val versionRegex = "([0-9]+)\\.([0-9]+).*".r - val hasExplainedVariance = metadata.sparkVersion match { - case versionRegex(major, minor) => - major.toInt >= 2 || (major.toInt == 1 && minor.toInt > 6) - case _ => false - } + val versionRegex = "([0-9]+)\\.(.+)".r + val versionRegex(major, _) = metadata.sparkVersion val dataPath = new Path(path, "data").toString - val model = if (hasExplainedVariance) { + val model = if (major.toInt >= 2) { val Row(pc: DenseMatrix, explainedVariance: DenseVector) = sparkSession.read.parquet(dataPath) .select("pc", "explainedVariance") .head() new PCAModel(metadata.uid, pc, explainedVariance) } else { - val Row(pc: DenseMatrix) = sparkSession.read.parquet(dataPath).select("pc").head() - new PCAModel(metadata.uid, pc, Vectors.dense(Array.empty[Double]).asInstanceOf[DenseVector]) + // pc field is the old matrix format in Spark <= 1.6 + // explainedVariance field is not present in Spark <= 1.6 + val Row(pc: OldDenseMatrix) = sparkSession.read.parquet(dataPath).select("pc").head() + new PCAModel(metadata.uid, pc.asML, + Vectors.dense(Array.empty[Double]).asInstanceOf[DenseVector]) } DefaultParamsReader.getAndSetParams(model, metadata) model --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@spark.apache.org For additional commands, e-mail: commits-h...@spark.apache.org
