repromodel: A Python repository from ReproModel

ReproModel

Model Comparison Tool for AI Researchers

ReproModel helps the AI research community to reproduce, compare, train, and test AI models faster.

ReproModel toolbox revolutionizes research efficiency by providing standardized models, dataloaders, and processing procedures. It features a comprehensive suite of pre-existing experiments, a code extractor, and an LLM descriptor. This toolbox allows researchers to focus on new datasets and model development, significantly reducing time and computational costs.

With this no-code solution, you'll have access to a collection of benchmark and SOTA models and datasets. Dive into training visualizations, effortlessly extract code for publication, and let our LLM-powered automated methodology description writer do the heavy lifting.

The current prototype helps researchers to modularize their development and compare the performance of each step in the pipeline in a reproducible way. This prototype version helped us reduce the time for model development, computation, and writing by at least 40%. Watch our demo.

The coming versions will help researchers build upon state-of-the-art research faster by just loading the previously published study ID. All code, experiments, and results will already be verified and stored in our system.

https://repromodel.netlify.app

Features

✅ Standard Models Included
✅ Known Datasets
✅ Metrics (100+)
✅ Losses (20+)
✅ Data Splitting
✅ Augmentations
✅ Optimizers (10+)
✅ Learning Rate Schedulers
✅ Early Stopping Criterion
✅ Training Device Selection
✅ Logging (Tensorboard ...)
✅ AI Experiment Description Generator
✅ Code Extractor
✅ Custom Script Editor
🔲 GUI augmentation builder
🔲 Docker image
🔲 Conventional ML models workflow
🔲 Parallel training
🔲 Statistical testing

Documentation

For examples and step-by-step instructions, please visit our full documentation at https://www.repromodel.com/docs.

Running Locally

You will need to have Node.js installed.

Combines npm install, creation of a virtual environment, as well as the launch of the frontend and backend:

npm run repromodel            // Mac and Linux
npm run repromodel-windows    // Windows

Launch the frontend instead individually:

npm install
npm run dev

If you want to use the Methodology Generator, you need to have Ollama installed You can get Ollama from their website and pull the model of your choice.

npm install
npm run repromodel-with-llm            // Mac and Linux
npm run repromodel-with-llm-windows    // Windows

Contributing

Contributions are what make the open-source community such an amazing place to learn, inspire, and create.

Any contributions you make are greatly appreciated. If you have a suggestion that would make this better, please read our Contribution Guidelines and Code of Conduct.