Metadata-Version: 2.4
Name: nnflowkit
Version: 3.2.4
Summary: A lightweight and comprehensive toolkit for end-to-end management of deep learning neural networks.
Author-email: nnflowkit <nnflowkit@nnflowkit.com>
License: MIT
Keywords: nnflowkit
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3.8
Classifier: Operating System :: OS Independent
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: <3.12,>=3.11
Description-Content-Type: text/markdown
Requires-Dist: torch>=2.1.2
Requires-Dist: torchvision>=0.16.2

# nnflowkit
A production-grade end-to-end management toolkit for deep learning neural networks, developed for researchers and engineering teams to streamline the entire neural network development workflow.

## Compatibility
⚠️ **Python Version Requirement**: nnflowkit is exclusively compatible with Python 3.10 (no support for other Python versions).

## Key Features
### 1. Neural Network Construction
- Support for modular building of complex network architectures, compatible with PyTorch ecosystem and custom layer definition.
- Native adaptation to spatiotemporal prediction scenarios, with pre-built modular components for time-series/spatial feature extraction.

### 2. Training Lifecycle Orchestration
- Unified interface for training process management: data loader integration, optimizer configuration, learning rate scheduling, and training progress tracking.
- Lightweight training loop encapsulation, balancing flexibility and ease of use for both rapid prototyping and production training.

### 3. Weight Management
- Automatic detection and saving of optimal training weights (based on validation metrics).
- Secure weight loading for network reconstruction, supporting weight versioning and cross-environment compatibility.

### 4. Inference & Deployment
- Optimized inference pipeline for trained networks, supporting batch inference and real-time single-sample inference.
- Seamless integration with PyTorch inference tools, ensuring consistency between training and inference results.

## Why nnflowkit?
- End-to-end coverage: Eliminates the need for fragmented tools for network building, training, weight management and inference.
- Ecosystem compatibility: Deep integration with PyTorch, consistent with mainstream deep learning development habits.
- Lightweight & efficient: No redundant dependencies, focused on core workflow optimization for neural network development.

## Installation
```bash
# Ensure Python 3.10 environment before installation
pip install nnflowkit
