feat(operator): give the KNN trainers' metric and metric_params a working converter - #7598
Open
kz930 wants to merge 1 commit into
Open
feat(operator): give the KNN trainers' metric and metric_params a working converter#7598kz930 wants to merge 1 commit into
kz930 wants to merge 1 commit into
Conversation
…king converter
SklearnAdvancedKNNParameters pairs each hyperparameter with the Python callable
that converts what the user typed. Two of the seven named one that cannot
produce what scikit-learn accepts, so picking either failed the run whatever was
entered, from the same dropdown as the five that work. Both trainers share the
enum, so both were affected.
metric was declared int. Its accepted values are words, so int("minkowski")
raises before scikit-learn sees anything, and a number that does convert is
rejected as not one of the accepted names. It is now str, which weights and
algorithm beside it already are.
metric_params takes a mapping of extra keyword arguments for the metric, and
none of int, float or str returns one, so a well-formed {"p": 2} arrived as that
same text. It now names json.loads. The type has never been limited to builtins
(SVC and SVR name an inline lambda for their boolean parameters), but json.loads
needs the module, so the generated template imports json. Unconditionally rather
than when such a parameter is present: the alternative is threading each
converter's imports through every ParamClass for one parameter of one operator.
Checked against scikit-learn both ways. Before: int("minkowski") raises
ValueError, and metric=int("3") reaches the estimator and is rejected as not an
accepted metric name. After: minkowski with metric_params {"p": 3} fits and
predicts with the mapping arriving as a dict, chebyshev fits on the regressor,
and mahalanobis with a VI matrix fits, which is what metric_params is for.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Contributor
Automated Reviewer SuggestionsBased on the
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #7598 +/- ##
=========================================
Coverage 88.63% 88.64%
- Complexity 4346 4347 +1
=========================================
Files 1177 1177
Lines 46896 46896
Branches 5230 5230
=========================================
+ Hits 41568 41569 +1
+ Misses 3568 3567 -1
Partials 1760 1760
*This pull request uses carry forward flags. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Contributor
|
| config | throughput | MB/s | latency | max Δ latest / 7d | |
|---|---|---|---|---|---|
| 🔴 | bs=10 sw=10 sl=64 | 693 | 0.423 | 15,336/17,511/17,511 us | 🔴 +26.5% / 🔴 +20.3% |
| 🔴 | bs=100 sw=10 sl=64 | 1,613 | 0.984 | 62,180/82,583/82,583 us | 🔴 +8.4% / 🟢 +61.2% |
| 🔴 | bs=1000 sw=10 sl=64 | 2,059 | 1.257 | 483,282/541,119/541,119 us | 🔴 +6.1% / 🟢 +100.7% |
Baseline details
Latest main 7daf8d7 from same runner
| config | metric | PR | latest main | 7d avg | Δ latest | Δ 7d |
|---|---|---|---|---|---|---|
| bs=10 sw=10 sl=64 | throughput | 693 tuples/sec | 849 tuples/sec | 775.33 tuples/sec | -18.4% | -10.6% |
| bs=10 sw=10 sl=64 | MB/s | 0.423 MB/s | 0.518 MB/s | 0.473 MB/s | -18.3% | -10.6% |
| bs=10 sw=10 sl=64 | p50 | 15,336 us | 12,124 us | 12,743 us | +26.5% | +20.3% |
| bs=10 sw=10 sl=64 | p95 | 17,511 us | 16,192 us | 16,310 us | +8.1% | +7.4% |
| bs=10 sw=10 sl=64 | p99 | 17,511 us | 16,192 us | 18,926 us | +8.1% | -7.5% |
| bs=100 sw=10 sl=64 | throughput | 1,613 tuples/sec | 1,732 tuples/sec | 1,001 tuples/sec | -6.9% | +61.2% |
| bs=100 sw=10 sl=64 | MB/s | 0.984 MB/s | 1.057 MB/s | 0.611 MB/s | -6.9% | +61.1% |
| bs=100 sw=10 sl=64 | p50 | 62,180 us | 57,381 us | 101,399 us | +8.4% | -38.7% |
| bs=100 sw=10 sl=64 | p95 | 82,583 us | 85,702 us | 108,206 us | -3.6% | -23.7% |
| bs=100 sw=10 sl=64 | p99 | 82,583 us | 85,702 us | 118,195 us | -3.6% | -30.1% |
| bs=1000 sw=10 sl=64 | throughput | 2,059 tuples/sec | 2,097 tuples/sec | 1,026 tuples/sec | -1.8% | +100.6% |
| bs=1000 sw=10 sl=64 | MB/s | 1.257 MB/s | 1.28 MB/s | 0.626 MB/s | -1.8% | +100.7% |
| bs=1000 sw=10 sl=64 | p50 | 483,282 us | 473,870 us | 996,304 us | +2.0% | -51.5% |
| bs=1000 sw=10 sl=64 | p95 | 541,119 us | 510,053 us | 1,042,531 us | +6.1% | -48.1% |
| bs=1000 sw=10 sl=64 | p99 | 541,119 us | 510,053 us | 1,074,934 us | +6.1% | -49.7% |
Raw CSV
config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,288.76,200,128000,693,0.423,15335.59,17510.87,17510.87
1,100,10,64,20,1240.04,2000,1280000,1613,0.984,62179.78,82582.90,82582.90
2,1000,10,64,20,9715.09,20000,12800000,2059,1.257,483282.44,541119.47,541119.47
Contributor
Author
|
@carloea2, @aglinxinyuan May you take a look at this? |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
What changes were proposed in this PR?
SklearnAdvancedKNNParameterspairs each hyperparameter with the Python callable that converts what the user typed. Two of the seven named one that cannot produce what scikit-learn accepts, so picking either failed the run whatever was entered, from the same dropdown as the five that work. Both trainers share the enum, so both were affected.metricwas declaredint. Its accepted values are words, soint("minkowski")raises before scikit-learn sees anything, and a number that does convert is rejected as not one of the accepted names. It is nowstr, whichweightsandalgorithmbeside it already are.metric_paramstakes a mapping of extra keyword arguments for the metric, and none ofint,floatorstrreturns one, so a well-formed{"p": 2}arrived as that same text. It now namesjson.loads. The type has never been limited to builtins, since the SVC and SVR trainers already name an inline lambda for their boolean parameters, butjson.loadsneeds the module, so the generated template imports json. Unconditionally rather than when such a parameter is present: the alternative is threading each converter's imports through everyParamClassfor one parameter of one operator.Any related issues, documentation, discussions?
Fixes #7593.
How was this PR tested?
int("minkowski")raises ValueError, andmetric=int("3")reaches the estimator and is rejected as not an accepted metric name. After: minkowski withmetric_params{"p": 3}fits and predicts with the mapping arriving as a dict, chebyshev fits on the regressor, and mahalanobis with a VI matrix fits, which is the casemetric_paramsexists for.WorkflowOperator/testOnly *sklearnAdvanced*: 7 suites, 40 tests, none failed.scalafmtCheckAlland both scalafix checks clean.Was this PR authored or co-authored using generative AI tooling?
Generated-by: Claude Code (Claude Opus 5)