Automated Tool for Hierarchical Exploration of Neural Architectures
Athena, named after the Greek goddess of wisdom and strategy, is an intuitive web-based tool designed to streamline the creation of machine learning models by generating starter code and meaningful visualizations. Whether you're a beginner looking for a quick way to structure your ML projects or an experienced practitioner prototyping architectures, Athena simplifies the process by providing ready-to-use code snippets and clear model diagrams.
https://athena-gamma-eight.vercel.app/
- 🔀 Model Types
- Support for Fully Connected (FCN) and Convolution Neural Networks (CNN)
- ⚙️ Model Architecture
- Add, remove, rearrange layers
- Edit layer parameters such as type, size, and other applicable parameters such as kernel and stride
- ✅ Validation
- Layer arrangement and all layer parameters are validated upon generation.
- ⚙️ Generation
- Single click generation of starter code and architecture visualization
- 🖼️ Interactive Visualization
- Interactive visualization with ability to expand into full screen, zoom and pan along with rotate for CNNs
- Clean and minimal architecture visualizations with configurable annotations settings
- 📋💾 Copy and Save
- Copy your code or download as a python file
- Download the model visualization as an image
- Choose the type of your model - FCN or CNN
- Add and configure layers and their parameters
- Move, delete and rearrange layers to achieve your desired model architecture
- Generate code and visualization with one click
- Configure visualization's annotation settings
- Download the visualization an an image
- Copy the code or save as a .py file
Currently the following frameworks/libraries are supported along with their respective paradigms for creating models
- PyTorch
- Keras
- Sequential
- Functional
- Subclassing
To run the project locally and set up a development environment, follow these steps:
Ensure you have the following installed:
git clone https://github.com/ChaoticTrio/athena.git
cd athenaUse npm to install the required packages:
npm installStart the local development server:
npm run dev- D3.js for FCN visualization
- Three.js for CNN visualization
- monaco-editor/react for Code editor