Implementation:Explodinggradients Ragas Validation Module

Field	Value
source	Explodinggradients_Ragas (GitHub)
domains	Validation, Framework
last_updated	2026-02-10 00:00 GMT

Overview

The validation module provides functions for validating that evaluation datasets have the required columns and that metrics are compatible with the dataset's sample type.

Description

remap_column_names renames columns in a HuggingFace Dataset using an inverse mapping, allowing datasets with non-standard column names to be used with Ragas metrics. get_supported_metric_type determines whether an EvaluationDataset contains SingleTurnSample or MultiTurnSample data and returns the corresponding MetricType name string. validate_required_columns checks that each metric's required columns are present in the dataset, raising a ValueError with a descriptive message listing missing columns. validate_supported_metrics verifies that each metric instance supports the dataset's sample type (single-turn or multi-turn), raising a ValueError if a metric does not match.

Usage

Call validate_required_columns and validate_supported_metrics before running an evaluation to ensure compatibility between the dataset and the selected metrics. Use remap_column_names to align column names when datasets use different naming conventions.

Code Reference

Item	Detail
Source Location	src/ragas/validation.py L14-63
Functions	remap_column_names(dataset, column_map), validate_required_columns(ds, metrics), validate_supported_metrics(ds, metrics), get_supported_metric_type(ds)
Import	from ragas.validation import validate_required_columns, validate_supported_metrics

I/O Contract

Inputs

Function	Parameter	Type	Description
remap_column_names	dataset	Dataset	HuggingFace Dataset to remap
remap_column_names	column_map	dict[str, str]	Mapping of standard names to dataset column names
validate_required_columns	ds	EvaluationDataset	Ragas evaluation dataset to validate
validate_required_columns	metrics	Sequence[Metric]	List of metrics to check column requirements for
validate_supported_metrics	ds	EvaluationDataset	Dataset to check sample type against
validate_supported_metrics	metrics	Sequence[Metric]	Metrics to validate compatibility for

Outputs

Function	Return Type	Description
remap_column_names	Dataset	Dataset with renamed columns
get_supported_metric_type	str	"SINGLE_TURN" or "MULTI_TURN"
validate_required_columns	None	Raises ValueError if columns are missing
validate_supported_metrics	None	Raises ValueError if metrics are incompatible

Usage Examples

from ragas.validation import (
    validate_required_columns,
    validate_supported_metrics,
    remap_column_names,
)
from ragas.dataset_schema import EvaluationDataset

# Validate dataset columns before evaluation
dataset = EvaluationDataset.from_list(my_samples)
metrics = [faithfulness_metric, answer_correctness_metric]

validate_required_columns(dataset, metrics)
validate_supported_metrics(dataset, metrics)

# Remap column names from custom dataset
from datasets import Dataset as HFDataset
raw_ds = HFDataset.from_dict({"query": [...], "answer": [...]})
column_map = {"user_input": "query", "response": "answer"}
remapped = remap_column_names(raw_ds, column_map)

Related Pages

• MetricValidators_Module - Value-level metric validation
• MetricResult_Class - Result objects validated by metric validators
• Utils_Module - General utility functions