Add docs for data/datasets and how to configure them in GuideLLM

[markurtz]

Fixes #133

[github-actions]

Build artifacts (.whl and .tar.gz) are available for download for up to 30 days.
They are located at https://github.com/neuralmagic/guidellm/actions/runs/14657660273/artifacts/3007230813

[github-actions]

Build artifacts (.whl and .tar.gz) are available for download for up to 30 days.
They are located at https://github.com/neuralmagic/guidellm/actions/runs/14657675243/artifacts/3007234607

[Copilot]

Pull Request Overview

This PR adds a new documentation page for configuring data and datasets in GuideLLM, and updates the README to include a link to the new guide.

• Introduces detailed guidance on dataset configurations, examples, and usage.
• Updates the main README to reference the new Data/Datasets Guide.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
docs/datasets.md	New documentation outlining dataset configuration and usage.
README.md	Updated navigation to include the new Data/Datasets guide.

[github-actions]

Build artifacts (.whl and .tar.gz) are available for download for up to 30 days.
They are located at https://github.com/neuralmagic/guidellm/actions/runs/14657701691/artifacts/3007241591

[SharonGil]

@markurtz Thank you very much for this, this simplifies and makes things much clearer.

2 small comments from my end:

1. In lines 27 and 142 - you put suffix .ext - just wanted to make sure this is intentionally as a place holder for some external file or a typo instead of .txt.
2. As of the data formats themselves, its written "Ensure the file format matches the expected structure for the dataset" - for HF model maybe to add a short elaboration regarding the 'dataset' library format expected, in order to understand which of HF models are supported and can be used and which aren't, rather than just trial and error...
   Also for the file based DS format - there are few examples there for .txt, .csv and .json files - do these represent the only formats supported for these files? Meaning - for .txt files the data should be plain newline delimited strings that represent the prompts without and keys added and without output, but for .csv for example you have to give a prompt and an output columns?

[markurtz]

@markurtz Thank you very much for this, this simplifies and makes things much clearer.

2 small comments from my end:

1. In lines 27 and 142 - you put suffix .ext - just wanted to make sure this is intentionally as a place holder for some external file or a typo instead of .txt.
2. As of the data formats themselves, its written "Ensure the file format matches the expected structure for the dataset" - for HF model maybe to add a short elaboration regarding the 'dataset' library format expected, in order to understand which of HF models are supported and can be used and which aren't, rather than just trial and error...
   Also for the file based DS format - there are few examples there for .txt, .csv and .json files - do these represent the only formats supported for these files? Meaning - for .txt files the data should be plain newline delimited strings that represent the prompts without and keys added and without output, but for .csv for example you have to give a prompt and an output columns?

@SharonGil take a look through again now, I've updated based on this feedback to make the points you raised clearer and try and answer your questions

[github-actions]

Build artifacts (.whl and .tar.gz) are available for download for up to 30 days.
They are located at https://github.com/neuralmagic/guidellm/actions/runs/14669119659/artifacts/3011014183

[SharonGil]

@markurtz That's perfect. Thanks a lot for the clarifications!