modularlibTrain

modularlibTrain is a Python library designed to facilitate the creation and management of modular deep learning models. It allows users to define and connect different neural network components, such as layers, blocks, and even entire sub-models, in a flexible and organized manner. The core concept revolves around a "Module" object, which represents a self-contained unit of computation that can be instantiated, connected to other modules, and combined to form larger architectures.

The library provides a base class for creating custom modules, enabling users to encapsulate specific functionalities and reuse them across different projects. These modules can have their own trainable parameters, forward pass logic, and internal states. Connections between modules are typically defined through a graph-like structure, where the output of one module can serve as the input to another. This compositional approach simplifies the design of complex models and promotes code reusability.

modularlibTrain aims to streamline the experimentation process by making it easier to swap out components, modify architectures, and manage dependencies. It offers utilities for loading and saving model configurations, as well as for visualizing the model's structure. The library is particularly useful for researchers and practitioners who frequently work with evolving model designs or require a more structured way to build and manage their deep learning pipelines.