You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@spark.apache.org by sr...@apache.org on 2019/01/23 01:22:24 UTC
[spark] branch master updated: [SPARK-26228][MLLIB] OOM issue
encountered when computing Gramian matrix
This is an automated email from the ASF dual-hosted git repository.
srowen pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/spark.git
The following commit(s) were added to refs/heads/master by this push:
new 6dcad38 [SPARK-26228][MLLIB] OOM issue encountered when computing Gramian matrix
6dcad38 is described below
commit 6dcad38ba3393188084f378b7ff6dfc12b685b13
Author: Sean Owen <se...@databricks.com>
AuthorDate: Tue Jan 22 19:22:06 2019 -0600
[SPARK-26228][MLLIB] OOM issue encountered when computing Gramian matrix
## What changes were proposed in this pull request?
Avoid memory problems in closure cleaning when handling large Gramians (>= 16K rows/cols) by using null as zeroValue
## How was this patch tested?
Existing tests.
Note that it's hard to test the case that triggers this issue as it would require a large amount of memory and run a while. I confirmed locally that a 16K x 16K Gramian failed with tons of driver memory before, and didn't fail upfront after this change.
Closes #23600 from srowen/SPARK-26228.
Authored-by: Sean Owen <se...@databricks.com>
Signed-off-by: Sean Owen <se...@databricks.com>
---
.../spark/mllib/linalg/distributed/RowMatrix.scala | 40 ++++++++++++++++++----
1 file changed, 34 insertions(+), 6 deletions(-)
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala
index ff02e5d..56caeac 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala
@@ -119,11 +119,25 @@ class RowMatrix @Since("1.0.0") (
val nt = if (n % 2 == 0) ((n / 2) * (n + 1)) else (n * ((n + 1) / 2))
// Compute the upper triangular part of the gram matrix.
- val GU = rows.treeAggregate(new BDV[Double](nt))(
- seqOp = (U, v) => {
+ val GU = rows.treeAggregate(null.asInstanceOf[BDV[Double]])(
+ seqOp = (maybeU, v) => {
+ val U =
+ if (maybeU == null) {
+ new BDV[Double](nt)
+ } else {
+ maybeU
+ }
BLAS.spr(1.0, v, U.data)
U
- }, combOp = (U1, U2) => U1 += U2)
+ }, combOp = (U1, U2) =>
+ if (U1 == null) {
+ U2
+ } else if (U2 == null) {
+ U1
+ } else {
+ U1 += U2
+ }
+ )
RowMatrix.triuToFull(n, GU.data)
}
@@ -136,8 +150,14 @@ class RowMatrix @Since("1.0.0") (
// This succeeds when n <= 65535, which is checked above
val nt = if (n % 2 == 0) ((n / 2) * (n + 1)) else (n * ((n + 1) / 2))
- val MU = rows.treeAggregate(new BDV[Double](nt))(
- seqOp = (U, v) => {
+ val MU = rows.treeAggregate(null.asInstanceOf[BDV[Double]])(
+ seqOp = (maybeU, v) => {
+ val U =
+ if (maybeU == null) {
+ new BDV[Double](nt)
+ } else {
+ maybeU
+ }
val n = v.size
val na = Array.ofDim[Double](n)
@@ -150,7 +170,15 @@ class RowMatrix @Since("1.0.0") (
BLAS.spr(1.0, new DenseVector(na), U.data)
U
- }, combOp = (U1, U2) => U1 += U2)
+ }, combOp = (U1, U2) =>
+ if (U1 == null) {
+ U2
+ } else if (U2 == null) {
+ U1
+ } else {
+ U1 += U2
+ }
+ )
bc.destroy()
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@spark.apache.org
For additional commands, e-mail: commits-help@spark.apache.org