drytorch 0.1.0b5

This package helps train, document and evaluate a Pytorch model.

DRYTorch

💡 Design Philosophy

By adhering to the Don't Repeat Yourself (DRY) principle, this library makes your machine-learning projects easier to replicate, document, and reuse.

✨ Features at a Glance

• Experimental Scope: All logic runs within a controlled scope, preventing unintended dependencies, data leakage, and misconfiguration.
• Modularity: Components communicate via defined protocols, providing type safety and flexibility for custom implementations.
• Decoupled Tracking: Logging, plotting, and metadata are handled by an event system that separates execution from tracking.
• Lean Dependencies: Minimal core requirements while supporting optional external libraries (Hydra, W&B, TensorBoard, etc.).
• Self-Documentation: Metadata is automatically extracted in a standardized and robust manner.
• Ready-to-Use Implementations: Advanced functionalities with minimal boilerplate, suitable for a wide range of ML applications.

📦 Installation

Requirements The library only requires recent versions of PyTorch and NumPy. Tracker dependencies are optional.

Installation Using pip:

pip install drytorch

Using uv

uv add drytorch

🗂️ Library Organization

The library uses a microkernel (plugin) architecture to separate concerns.

1. Core (core): The library kernel. Contains the Event System, Protocols for component communication, and internal safety Checks.
2. Standard Library (lib): Reusable implementations and abstract classes of the protocols.
3. Trackers (tracker): Optional tracker plugins that integrate via the event system.
4. Contributions (contrib): Dedicated space for community-driven extensions.
5. Utilities (utils): Functions and classes independent to the framework.

📙 Notebook Tutorials

Dive into the full, runnable examples:

➡️ Getting Started Notebook

➡️ Metrics and Losses Notebook

➡️ Experiments and Runs Notebook

📝 Changelog