Add entry in dataset card to help fine-tuning using TRL with the generated dataset

[plaguss]

Description

• Add PRM for MathShepherd tasks.

This PR adds a new method to the _Step class called _dataset_use. By default doesn't contain anything, but can be used to enhance the dataset card of a Distiset. The examples implemented here correspond to the classes that prepare the data for fine-tuning, it includes an example command using TRL.

To include a new "use" for a dataset associated with a step, we have to override the _dataset_use method, take a look at FormatTextGenerationSFT or FormatTextGenerationDPO for a example.

An example can be seen in this dummy dataset: https://huggingface.co/datasets/plaguss/test_dataset_use

[github-actions]

Documentation for this PR has been built. You can view it at: https://distilabel.argilla.io/pr-1079/

[codspeed-hq]

CodSpeed Performance Report

Merging #1079 will improve performances by ×7.2

_{Comparing dataset-card-help (ca07a58) with develop (63c75c5)}

Summary

⚡ 1 improvements

Benchmarks breakdown

	Benchmark	develop	dataset-card-help	Change
⚡	test_cache_time	3,985.3 ms	553.1 ms	×7.2