feat: Add multi-model LLM benchmarking notebook#384
Open
xeroxpro wants to merge 2 commits intoed-donner:mainfrom
Open
feat: Add multi-model LLM benchmarking notebook#384xeroxpro wants to merge 2 commits intoed-donner:mainfrom
xeroxpro wants to merge 2 commits intoed-donner:mainfrom
Conversation
There was a problem hiding this comment.
Copilot wasn't able to review any files in this pull request.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Owner
|
Hey - this is great - but would you be OK to clear the outputs first for me? Otherwise it's 2,000+ loc |
…ols - Complete investment analysis system with multiple specialized agents - Financial, technical, fundamental, and sentiment analysis tools - ESG, geopolitical, and supply chain intelligence - Turkey economic intelligence and institutional analysis - Market data integration with yfinance - Portfolio optimization and risk assessment - Dashboard UI and report generation - Knowledge base builder and data fetcher utilities
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR introduces a Jupyter Notebook (.ipynb) that provides an end-to-end framework for evaluating and benchmarking multiple Large Language Models from various providers.
What this notebook contains:
Dynamic Client Initialization: The notebook automatically detects available API keys (OpenAI, Groq, DeepSeek) from environment variables and checks for local services (Ollama) to dynamically configure the list of competitor models.
Weighted Question Generation: Implements a two-step process to generate the evaluation questions:
Generates three distinct, nuanced questions.
Uses a "Consultant" LLM (gpt-4o-mini) to rank these questions and assign weights (50, 30, 20 points).
Parallel Query Execution: A loop iterates through all configured models and collects their responses to all three weighted questions.
Dual-Weight "Judge" Scoring:
All responses are formatted and sent to a single "Judge" LLM (gpt-4o).
The Judge provides two scores for each answer: a judge_score (60% weight) and a peer_average_score (40% weight).
Final Score Calculation: A script parses the Judge's nested JSON output, applies the answer-level (60/40) weights, and then applies the question-level (50/30/20) weights to calculate a final score for each model.
Results Visualization: Uses matplotlib to generate a horizontal bar chart of the final results, highlighting the winning model in a different color.
Dependencies:
The notebook requires the following Python packages:
openai
requests (for the Ollama health check)
matplotlib (for the final graph)
numpy
jupyter