[SPARK-12685] [MLlib] [Backport to 1.4]word2vec trainWordsCount gets overflow

[hhbyyh]

jira: https://issues.apache.org/jira/browse/SPARK-12685

master PR: #10627

the log of word2vec reports
trainWordsCount = -785727483
during computation over a large dataset.

Update the priority as it will affect the computation process.
alpha = learningRate * (1 - numPartitions * wordCount.toDouble / (trainWordsCount + 1))

[SparkQA]

Test build #49223 has finished for PR 10721 at commit 27ba586.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[srowen]

LGTM

[hhbyyh]

Hi @srowen , Thanks for taking a look. I'm not sure if I should send separate PRs for release 1.5 and 1.6, can you please advice? Thanks.

[srowen]

@hhbyyh do you know if it cherry-picks cleanly into other branches? @jkbradley indicated it didn't. Back-porting to 1.6 makes sense; 1.5 maybe; 1.4 seems pretty old as it's very unlikely to see another release.

[jkbradley]

@hhbyyh I think this PR will cherry-pick cleanly to 1.4, 1.5, and 1.6. I think it's a change in the last line in master (after 1.6) which messed up the original PR. The main benefit of a separate PR would be getting Jenkins to run tests, but I think this PR is pretty safe.

LGTM

I'll try merging now. @srowen I agree about 1.4, but I might as well if it cherry-picks cleanly.

[jkbradley]

The cherry-pick worked, so this is in 1.4, 1.5, 1.6.

[JoshRosen]

@hhbyyh, now that this has been merged would you mind closing this pull request? GitHub can't auto-close PRs which were opened against maintenance branches.

[hhbyyh]

Thanks. Close the PR now.