zachpinto / sota-explorer

Streamlit app for viewing detailed information about SOTA ML models and performances on popular benchmarks

SOTA Explorer

Description

This project creates a streamlit app to allow users to quickly find and view detailed information about ML tasks and how various models have performed on popular benchmarks. The app is designed to be user-friendly and easy to navigate. One theoretical use case is for journalists to quickly find a layman description of a given ML task and, if necessary, report on how specific capabilities related to that task (or its subtasks) have improved over time.

Features

• Search for SOTA models by task, dataset, or model name in a "nested" directory structure
• View detailed information about a task, including:
  • A layman's description of the task
  • A list of subtasks (if applicable)
  • A list of datasets used to benchmark the task
  • A chart showing the performance of various models on selected task over time

Data Sources

• Papers with Code for all ML tasks, datasets, and model evaluation metrics

Installation

Clone the repository and navigate to the project directory:

git clone https://github.com/zachpinto/SOTA-explorer.git
cd SOTA-Explorer

Install the required packages:

pip install -r requirements.txt

Usage

To run the application, execute the following command:

streamlit run HOME.py

Contributing

Pull requests are welcome. I am not a professional web developer, so major changes to other front-end technologies are welcome.

License

MIT