# mlflow
pip install git+https://github.com/mlflow/mlflow.git@refs/pull/19376/merge
# mlflow-skinny
pip install git+https://github.com/mlflow/mlflow.git@refs/pull/19376/merge#subdirectory=libs/skinny

%sh curl -LsSf https://raw.githubusercontent.com/mlflow/mlflow/HEAD/dev/install-skinny.sh | sh -s pull/19376/merge

Resolve #19374

What changes are proposed in this pull request?

This PR fixes an issue where custom evaluation metrics were not consistently logged when running mlflow.evaluate with multiple evaluators enabled. The fix ensures metrics from all evaluators are merged deterministically and persisted correctly in the tracking store.

How is this PR tested?

• Existing unit/integration tests
• New unit/integration tests
• Manual tests

Manual testing details

• Ran mlflow.evaluate with multiple evaluators on a sample classification model.
• Verified that all metrics appear in the run UI and via the Tracking API.
• Confirmed no regression in single-evaluator workflows.

Does this PR require documentation update?

• No. You can skip the rest of this section.

• Yes. I've updated:

  • Examples
  • API references
  • Instructions

Release Notes

Is this a user-facing change?

• Yes. Give a description of this change to be included in the release notes for MLflow users.

Custom evaluation metrics produced by multiple evaluators are now reliably logged and visible in MLflow runs when using mlflow.evaluate.

What component(s), interfaces, languages, and integrations does this PR affect?

Components

• area/tracking
• area/models
• area/model-registry
• area/scoring
• area/evaluation
• area/gateway
• area/prompts
• area/tracing
• area/projects
• area/uiux
• area/build
• area/docs

How should the PR be classified in the release notes? Choose one:

• rn/none
• rn/breaking-change
• rn/feature
• rn/bug-fix
• rn/documentation

Should this PR be included in the next patch release?

• Yes (this PR will be cherry-picked and included in the next patch release)
• No (this PR will be included in the next minor release)