Chartorch is a web-based visual interface that simplifies the creation, training, and testing of machine learning models. By leveraging an intuitive flowchart system, users can design PyTorch neural networks visually without writing boilerplate code, and immediately train or test them on simple datasets. Generated code can be inspected and copied.
Built with Flask, the application is designed with security in mind:
all user-generated model training and testing code is strictly sandboxed
and executed within an isolated nsjail environment.
- Visual Model Builder: Create PyTorch architectures using a drag-and-drop flowchart interface.
- Integrated Training & Testing: Train your custom models on standard datasets directly within the browser.
- Secure Execution: Utilizes
nsjailto safely isolate user-submitted execution processes from the host system.
Before installing Chartorch, ensure your system has the following:
- Linux environment (Required for
nsjailprocess isolation) - nsjail: Must be installed and accessible on your system path.
- Python 3.14 (or a compatible Python 3.x version)
- Clone the repository:
git clone https://github.com/brandonbate/chartorch.git cd chartorch - Create and activate a virtual environment:
python -m venv venv source venv/bin/activate - Install the required dependencies:
pip install -r requirements.txt
To run the application locally:
python app.pyOnce the server initializes, open your web browser and navigate to: http://localhost:5000