nn.Module
✦₊⁺ Overview
The nn.Module class is the foundation for all AI components in msgFlux, inspired by torch.nn.Module.
It provides a structured way to build, compose, and manage AI workflows with features like parameter serialization, hooks, and async support.
1. Quick Start
import msgflux.nn as nn
class MyWorkflow(nn.Module):
def __init__(self):
super().__init__()
self.register_buffer("greeting", "Hello!")
def forward(self, name: str) -> str:
return f"{self.greeting} {name}"
workflow = MyWorkflow()
result = workflow("World") # "Hello! World"
print(result)
print(workflow.state_dict())
2. Built-in Modules
msgFlux provides ready-to-use modules:
| Module | Description |
|---|---|
nn.Sequential |
Chain modules in sequence |
nn.ModuleDict |
Dictionary of named modules |
nn.ModuleList |
Ordered list of modules |
nn.Agent |
LM with tools and reasoning |
nn.Transcriber |
Speech-to-text |
nn.Speaker |
Text-to-speech |
nn.Searcher |
Data retrieval |
nn.Embedder |
Text embeddings |
nn.MediaMaker |
Image/video generation |
nn.Predictor |
ML model wrapper (sklearn, etc.) |
4. Contents
| Topic | Description |
|---|---|
| Core Concepts | Parameters, buffers, and state dict |
| Forward and Async | Execution methods and hooks |
| Composing Modules | Sub-modules and nested composition |
| ModuleDict | Dictionary of named modules |
| ModuleList | Ordered list of modules |
| Sequential | Chain of modules |
| Visualization | Flow diagrams and complete example |