Skip to content

[SPARK-30503][ML] OnlineLDAOptimizer does not handle persistance correctly #27261

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
zhengruifeng wants to merge 1 commit into from
Closed

[SPARK-30503][ML] OnlineLDAOptimizer does not handle persistance correctly #27261

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion mllib/src/main/scala/org/apache/spark/mllib/clustering/LDAOptimizer.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,7 +142,7 @@ final class EMLDAOptimizer extends LDAOptimizer {
// For each document, create an edge (Document -> Term) for each unique term in the document.
val edges: RDD[Edge[TokenCount]] = docs.flatMap { case (docID: Long, termCounts: Vector) =>
// Add edges for terms with non-zero counts.
termCounts.asBreeze.activeIterator.filter(_._2 != 0.0).map { case (term, cnt) =>
termCounts.nonZeroIterator.map { case (term, cnt) =>
Edge(docID, term2index(term), cnt)
}
}
Expand Down Expand Up @@ -211,11 +211,14 @@ final class EMLDAOptimizer extends LDAOptimizer {
val docTopicDistributions: VertexRDD[TopicCounts] =
graph.aggregateMessages[(Boolean, TopicCounts)](sendMsg, mergeMsg)
.mapValues(_._2)
val prevGraph = graph
Copy link
Member

@srowen srowen Jan 17, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I buy it, but where is graph persisted again?

Copy link
Member

@srowen srowen Jan 22, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@zhengruifeng I'm still kind of curious, but trust your judgment so I will merge it.

Copy link
Contributor Author

@zhengruifeng zhengruifeng Jan 23, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! I am on the holidy, so can no reply quickly.
Current change is following what Pregel does,
I will go on looking into it, since the persistance in GraphX is a little complex.

// Update the vertex descriptors with the new counts.
val newGraph = Graph(docTopicDistributions, graph.edges)
graph = newGraph
graphCheckpointer.update(newGraph)
globalTopicTotals = computeGlobalTopicTotals()
prevGraph.unpersistVertices()
prevGraph.edges.unpersist()
this
}

Expand Down