Convert TopK reduction to be multiline example based #1752
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
JohnLangford
reviewed
Feb 8, 2019
| weighted label sum = 15.000000 | ||
| average loss = 0.000002 | ||
| best constant = 1.666667 | ||
| total feature number = 39 | ||
| total feature number = 36 |
There was a problem hiding this comment.
This changes because you eliminate newline examples with a constant feature?
JohnLangford
reviewed
Feb 8, 2019
| 0.000004 0.000001 6 6.0 1.0000 1.0007 4 | ||
| 0.000003 0.000000 7 7.0 2.0000 2.0004 4 | ||
| 0.000003 0.000000 8 8.0 1.0000 1.0003 4 | ||
| 0.000002 0.000000 9 9.0 3.0000 2.9995 4 |
There was a problem hiding this comment.
The number of multiline examples should be 3, right?
There was a problem hiding this comment.
Yes it is 3, but I am recreating the previous behavior of printing statistics after each example. See line 91 in the reduction. What do you think is the correct thing to do here? Only print the statistics after each set of examples?
There was a problem hiding this comment.
Yes. All predictions from a top-k call should (conceptually) be on a single line. The idea of this reduction is that you are return the top-k-of-n things.
|
Merged, thanks :-) |
rajan-chari
pushed a commit
to VowpalWabbit/reinforcement_learning
that referenced
this pull request
Feb 22, 2019
RunTests: use test label number instead of counter (VowpalWabbit/vowpal_wabbit#1753) Small Json parser cleanup (VowpalWabbit/vowpal_wabbit#1759) Type erase json parser context for easier deletion ((VowpalWabbit/vowpal_wabbit#1760) Fix static linking (VowpalWabbit/vowpal_wabbit#1758) Fix build scripts forcing Debug builds. Add LTO mode and fix VW defau…(VowpalWabbit/vowpal_wabbit#1735) Do not define BOOST_TEST_DYN_LINK when statically linking (VowpalWabbit/vowpal_wabbit#1750) Convert TopK reduction to be multiline example based (VowpalWabbit/vowpal_wabbit#1752) vw java 11 compatibility (VowpalWabbit/vowpal_wabbit#1700) cbify: --cbify_ldf for multiline (csoaa_ldf) input datasets (VowpalWabbit/vowpal_wabbit#1681) Merge pull request #1751 from yannstad/fix-tests (VowpalWabbit/vowpal_wabbit#1751) [tests] Make repeat.py compatible with python 3 (VowpalWabbit/vowpal_wabbit#1747)
jackdoe
pushed a commit
to lucjb/vowpal_wabbit
that referenced
this pull request
Mar 7, 2019
* Make topk a multiline learner * Fix test for new format and rename B to K * revert destructor usage * remove all from data, move to output seq in finish * Revert "remove all from data, move to output seq in finish" This reverts commit 2c8f965.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #1748