[tarun-google]

The Java and Go RunInference cross-language tests hardcode where the Python SDK
harness puts staged files (<semiPersistDir>/staged/sklearn_model, or
/tmp/staged/sklearn_model in Go). That location belongs to the container's
boot program and is only known on the worker, so the tests break when it
changes (see #40699, which moves Python staging to
<semi_persist_dir>/<worker id>/staged).

The test expansion service (expansion_service_test.py) attaches the model to
the Python environment as a staged file named sklearn_model. The tests now
pass that name and use a small SklearnModelHandlerNumpy subclass that builds
the full path from SEMI_PERSISTENT_DIRECTORY in load_model(). load_model()
runs on the worker after boot has set that variable. The subclass uses the
public update_model_path(), can be called repeatedly, and raises a KeyError
naming the variable if it is missing.

Test-only change. It works with both the current and the per-worker staging
layout, so it can merge ahead of #40699.

Testing (DirectRunner + Python expansion service, Python 3.11):
./gradlew :sdks:python:test-suites:direct:xlang:validatesCrossLanguageRunnerJavaUsingPython --tests '*RunInferenceTransformTest*' -PpythonVersion=3.11

[github-actions]

📋 Review Summary

This PR updates the Java and Go cross-language RunInference tests to resolve the staged sklearn_model path at worker runtime via SEMI_PERSISTENT_DIRECTORY in load_model(), rather than hardcoding the staging directory at pipeline construction time. The implementation is clean, idempotent across multiple load_model() calls, and ensures compatibility with per-worker staging directory layouts (#40699).

🔍 General Feedback

• Using os.path.basename(self._model_uri) combined with self.update_model_path(...) cleanly keeps load_model() idempotent if called multiple times.
• Both the unkeyed and keyed (KeyedModelHandler) test paths in Java as well as the Go xlang integration test are properly covered.

---

💡 Tip: Comment /gemini review on this pull request to trigger a new Gemini code review.

[github-actions]

Checks are failing. Will not request review until checks are succeeding. If you'd like to override that behavior, comment assign set of reviewers