Skip to content

Conversation

@AnthonyTruchet
Copy link

@AnthonyTruchet AnthonyTruchet commented Nov 18, 2016

What changes were proposed in this pull request?

The zero for the aggregation used to be shipped into a closure which is
higly problematic when this zero is big (100s of MB is typical for ML).
This change introduces a new overload of treeAggregate which only ships a
function able to generate this zero.

NB This review takes into account preliminary feedback from #15905

How was this patch tested?

Unit tests for core module launched locally

Licensing

This contribution is made on behalf of CRITEO (http://labs.criteo.com/) under the terms of the Apache 2.0 license.

CostFun used to send a dense vector of zeroes as a closure in a
treeAggregate call. To avoid that, we replace treeAggregate by
mapPartition + treeReduce, creating a zero vector inside the mapPartition
block in-place.
@AmplabJenkins
Copy link

Can one of the admins verify this patch?

@srowen
Copy link
Member

srowen commented Nov 18, 2016

See my comments from your previous PR; they still apply. #15905

@AnthonyTruchet
Copy link
Author

Sure, sorry I mistakenly pushed a badly rebased version

@AnthonyTruchet AnthonyTruchet deleted the ENG-17719-lbfgs-only branch November 21, 2016 12:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants