torchboard 0.1.2

Python dashboard application adding interactivity into your pytorch model

Torch-Board

Torchboard is a powerful library for real-time neural network modification and monitoring in PyTorch. Torchboard provides an interactive dashboard to tweak model architectures, visualize training metrics, and optimize hyperparameters on the fly.

Installation

Install Torch-Board via pip:

pip install torchboard 

🧰 Features

• 📊 Live Metrics Dashboard: Visualize loss, accuracy, gradients, and other metrics in real time through an intuitive interface.
• 🛠️ Interactive Model Editing: Modify layers, activations, hyperparameters, and even stop or adjust training without restarting the training loop.
• 🌐 Automatic Localhost Hosting: The dashboard is automatically hosted on localhost, enabling seamless interaction with your model via a web interface.
• 🔍 Advanced Visualization Tools: Monitor weight distributions, activations, computational graph changes, and other insights like learning rate, accuracy, and loss curves.
• 🤝 Seamless PyTorch Integration: Supports most PyTorch optimizers and integrates effortlessly into existing workflows.

Contributing

Contributions are welcome! Please submit a pull request or open an issue to add new decorators or suggest improvements.

License

This project is licensed under the Apache v2.0 License.