Repository navigation
silhouette_samples/silhouette_score: metric_params collides with pinned arguments #713
Description
Problem
Passing metric_params={"be": ...} (or **kwds with be) to tslearn.clustering.silhouette_samples raises TypeError: got multiple values for keyword argument 'be', because the kwds are merged into metric_params_ and splatted into a call that already pins be=be.
Suggested fix
Treat be like n_jobs/verbose: strip it from metric_params_ (the backend is already pinned by instantiate_backend(X)), or explicitly reject it with a clear message. At minimum, document that be cannot be forwarded.
Metadata
Severity: P2
Confidence: 100
Reviewer(s): adversarial (ce-code-review run 20260821-221628-4e205155 on PR #703)
Finding ID: tslearn/clustering/utils.py:347 metric_params={'be': ...} collides with the backend argument
Activity
charavelg commentedon Aug 26, 2026
I guess the same is true fro silhouette_score, with the addition of the verbose param not being properly handled in silhouette_score.
Also
metric_paramsparameter documentations of both function need to be updated accordingly.Would be nice to pack this into a single PR.
[-]silhouette_samples: metric_params containing 'be' collides with pinned backend argument[/-][+]silhouette_samples/silhouette_score: metric_params collides with pinned arguments[/+]WilliamK112 commentedon Aug 26, 2026
I would like to take this and include the matching
silhouette_scorebehavior noted above in one focused PR. I will strip backend/parallelism parameters that are already pinned by the public arguments before forwarding metric parameters, document the override behavior for both functions, and add regression tests for bothmetric_paramsand direct keyword input. I will leave the separate precomputed behavior from #714 unchanged.