[SPARK-11847] [ML] Model export/import for spark.ml: LDA

[hhbyyh]

Add read/write support to LDA, similar to ALS.

save/load for ml.LocalLDAModel is done.
For DistributedLDAModel, I'm not sure if we can invoke save on the mllib.DistributedLDAModel directly. I'll send update after some test.

[SparkQA]

Test build #46492 has finished for PR 9894 at commit 58abcab.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * class LocalLDAModelWriter(instance: LocalLDAModel) extends MLWriter\n * class DistributedLDAModelWriter(instance: DistributedLDAModel) extends MLWriter\n

[jkbradley]

Reviewing now

[jkbradley]

scala style: put vocabSize on next line

[jkbradley]

Thanks for the PR! My only other comment is about the missing DistributedLDAModel unit test.

In the future, I also want to add an object LDA which implements Readable and returns an LDAModel. But we can skip that for now.

[hhbyyh]

@jkbradley Thanks for review.

Do you mean a LDAModel.load(path) which can load both local and distributed LDAModel? We can create a jira.

[SparkQA]

Test build #46581 has finished for PR 9894 at commit 44c8193.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):\n * class LocalLDAModelWriter(instance: LocalLDAModel) extends MLWriter\n * class DistributedWriter(instance: DistributedLDAModel) extends MLWriter\n

[jkbradley]

Do you mean a LDAModel.load(path) which can load both local and distributed LDAModel? We can create a jira.

Yes, that's what I had in mind. That way, most users will never need to know about local vs. distributed.

[jkbradley]

LGTM. @hhbyyh Thanks for the updates! Is there anything left to do? If not, can you please remove the "WIP" label? I can then merge it.

[hhbyyh]

@jkbradley removed. Thanks. I'll create a jira.

[mengxr]

Merged into master and branch-1.6. Thanks!