API Reference / agent / tools / workflow /

benchmark.py

Multi-algorithm comparison.

Functions

Func

execute_benchmark

Line 9
execute_benchmark(**kwargs) -> Dict[str, Any]

Execute a model comparison via tuiml.Benchmark.

Backs the tuiml_benchmark tool: runs every algorithm on every dataset with cross-validation and aggregates per-fold scores.

Parameters

data
str or list of str
Dataset name(s) to benchmark on: dataset_ids, file paths, or built-in names (arrives via **kwargs, like all parameters below).
algorithms
list
Algorithms to compare; entries are names or {"name": ..., "params": {...}} dicts.
cv
int = 10
Number of cross-validation folds.
metrics
list of str = None
Metrics to compute; defaults to the benchmark's auto selection.
random_seed
int = None
Random seed for reproducible folds.
_progress_callback
callable = None
Internal per-fold progress hook; stripped from recorded args.

Returns

result
dict
On success: status ('success'), summary (text), table_markdown, results (nested {dataset: {model: {metric: {mean, std, scores}}}}), algorithms, datasets, cv_folds, random_seed, and optionally progress_log and research_log_updates. On failure: status ('error') and error.