---

name: grid-search description: "Run automl_gs / automl_grid_search on a CSV and target field." disable-model-invocation: true metadata: disco-role: operating license: MIT

Grid Search

Use this sub-skill to launch a bounded automl-gs search on a local CSV, choose a framework, infer or override the target metric, and inspect the trial log.

Route here for

• automl_gs CLI and automl_grid_search(...)
• framework selection (tensorflow, xgboost)
• target-metric inference, override, and result selection
• trial count, epoch budget, and output-folder discovery
• smoke checks for import/signature, CLI help, and tiny offline XGBoost search

Do not handle here

• Post-search use of generated model.py, pipeline.py, encoder JSON, or retraining/prediction on the exported artifacts; route to generated-artifacts.

Operating workflow

1. Read api-reference for install/import notes, exact call signatures, and typing overrides.
2. Read hyperparameters-and-metrics for search-space rules, metric direction, and output layout.
3. Use troubleshooting when imports, dependencies, type inference, or CSV paths go wrong.
4. For a fast smoke run, start from scripts/run_tiny_xgboost_search.py.

Keep this sub-skill focused on search-time behavior. Handle exported artifacts elsewhere.