Files
Tobias Pitters c3481c11aa Refactor short circuiting of tree_path_dependent shap calculation (#5069)
* refactor internal short circuiting for internal tree model shap calculation

* refactor

* remove unnecessary if statement

* add version check for categorical xgboost handling

* regenerate images

* update selenium-generated images

* limit numpy so that more recent version of numba is installed

* fix mypy tests, unpin gpboost

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* ignore mypy error for cutils

* add actions to all

* Make Selenium force-plot baseline comparison tolerant to Chrome version drift

Chrome/Chromedriver versions differ between local machines and CI runners
(both unpinned), which can shift the rendered screenshot size by a few
pixels with no actual change to the plot, causing spurious failures. Allow
a small size tolerance and diff over the shared cropped region instead of
requiring an exact dimension match.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* revert to a working state for maskers, ignore the line in mypy

* fix numba issue reference in pyproject.toml

* add graphviz to fix failing notebooks

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
2026-07-20 14:59:00 +02:00
..
2026-04-17 22:35:19 +02:00
2026-04-17 22:35:19 +02:00
2025-10-14 08:52:59 +02:00

SHAP Tests

GPU TreeExplainer Tests

The GPU TreeExplainer tests require a CUDA-capable GPU and the _cext_gpu extension built from source. Since GPU hardware is not available in our CI pipeline, these tests can be run manually on Google Colab.

Running on Google Colab

Open In Colab

  1. Click the badge above to open the notebook in Google Colab
  2. Switch to a GPU runtime: Runtime > Change runtime type > T4 GPU
  3. Run all cells