[SC-9395] Add RAG benchmarking demo notebook

[juanmleng]

Internal Notes for Reviewers

Added titles to RAGAS metrics because when using comparison tests, figures without titles displayed "None" in the title updated by the comparison logic.

External Release Notes

Added a new notebook rag_benchmarking_demo.ipynb to demonstrate benchmarking of the RAG RFP use case by comparing multiple configurations at each stage of the pipeline. It evaluates two embedding models (OpenAI's text-embedding-3-small and text-embedding-3-large), two retrieval models with different k parameters (5 and 10), and two LLM generators (gpt-3.5-turbo and gpt-4o), creating a total of four complete RAG pipelines. The notebook runs the same tests included in rag_documentation_demo.ipynb such as RAGAS metrics (Context Precision, Faithfulness, Answer Correctness), generation quality metrics (ROUGE, BLEU, BERT Score), and bias/toxicity evaluations on all configurations.

[johnwalz97]

lgtm

[github-actions]

PR Summary

This pull request introduces a comprehensive notebook for benchmarking Retrieval-Augmented Generation (RAG) models using the ValidMind library. The notebook demonstrates setting up a RAG pipeline with multiple embedding, retrieval, and generation models, including OpenAI's GPT-3.5 and GPT-4o. It showcases how to initialize models, load datasets, perform data validation, and run various tests to assess model performance. Additionally, the PR enhances visualization in several RAGAS tests by adding titles to histograms and box plots, improving the clarity of the visual output.

Test Suggestions

• Run the notebook end-to-end to ensure all cells execute without errors.
• Verify that the RAG pipeline correctly initializes and processes the dataset.
• Check the output of the data validation tests for expected results.
• Ensure that the embedding, retrieval, and generation models produce expected outputs.
• Validate that the visualizations in the RAGAS tests display correctly with the new titles.
• Test the notebook with different datasets to ensure robustness.
• Check the compatibility of the notebook with different versions of the ValidMind library.