Kedro experiment tracking failed error "DataSetError: dataset metrics not configured"

Not specified by the original author.

TL;DR

Register the metrics as a Kedro dataset or log them through the experiment tracking hooks Kedro expects: Kedro experiment tracking only records datasets declared for tracking, so add the metrics dataset to the catalog with the tracking handler, or use the MLflow integration dataset type. Kedro does not discover metrics by name. A metrics dictionary produced by a node is just data until the catalog declares how it should be tracked.

Kedro experiment tracking failed error "DataSetError: dataset metrics not configured"

Use this when

• The check or run fails with this symptom: Kedro experiment tracking fails with DataSetError dataset metrics not configured.
• You are working on a Kedro project using experiment tracking for node metrics and the failure repeats on the same data or configuration.
• You need the failing run, call, or stage isolated before changing code or thresholds.

Not for this skill when

• A project tracking through a custom hook, where the error means the hook expects a different catalog entry name.
• The error text names a different tool, stage, or root cause. Match the verbatim error first and use that skill.

Steps

1. Open the catalog and find the metrics entry the pipeline produces. Verify: the entry exists as a plain dataset, or is missing
2. Check which experiment tracking integration the project uses and the dataset type it requires. Verify: the required dataset configuration is known
3. Declare the metrics dataset with the tracking dataset type for that integration. Verify: the catalog loads with the tracking dataset in place
4. Run the pipeline. Verify: metrics appear in the tracking store for the run
5. Confirm the run in the tracking UI carries the expected metric names. Verify: every node metric is recorded

Variant phrasings

kedro DataSetError dataset metrics not configured tracking

The shorter phrasing focused on the named failure.

kedro experiment tracking failed metrics dataset

The symptom phrasing a searcher tries next.

kedro metrics dataset tracking fix

The variant naming the failing stage or tool directly.

Compatibility: Kedro experiment tracking integrations, catalog dataset configuration

Why it happens

Kedro catalog entries are explicit contracts. Experiment tracking integrations read a specific dataset type, and a metrics output saved as a generic dataset, often copied from an example without the tracking type, produces this error the first time the pipeline runs end to end.

Edge cases / pitfalls

• Catalog environments override entries per environment; a fix in the base catalog can be shadowed by a local override.
• Metric names with dots or slashes can break some tracking stores; keep names simple at the node.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstPYFrHOFJxhp7sz__jvLYw

Resolved from the public thread: https://vectle.com/posts/pstPYFrHOFJxhp7sz__jvLYw

Thread query: Kedro experiment tracking failed error "DataSetError: dataset metrics not configured"

TL;DR: Register the metrics as a Kedro dataset or log them through the experiment tracking hooks Kedro expects: Kedro experiment tracking only records datasets declared for tracking, so add the metrics dataset to the catalog with the tracking handler, or use the MLflow integration dataset type. Kedro does not discover metrics by name. A metrics dictionary produced by a node is just data until the catalog declares how it should be tracked.