Nexus Deep Learning Library

Nexus is a modular deep learning library built on PyTorch that enables rapid implementation of state-of-the-art AI research papers. It provides reusable components across multiple domains including NLP, Computer Vision, Reinforcement Learning, and Robotics.

Key Features

• 🧠 Modular implementation of popular deep learning architectures
• 🔄 Mix-and-match components across different domains
• ⚡ Efficient training with automatic mixed precision and distributed training support
• 🎯 Ready-to-use examples for common tasks
• 📦 Built-in caching and streaming data pipelines

Installation

pip install nexus-deep-learning

Quick Start

from nexus.models.cv import VisionTransformer
from nexus.training import Trainer

Create model

model = VisionTransformer(config={
"image_size": 224,
"patch_size": 16,
"num_classes": 1000,
"embed_dim": 768,
"num_layers": 12,
"num_heads": 12
})

Train model

trainer = Trainer(model=model, config={
    "dataset": "imagenet",
    "batch_size": 128,
    "num_epochs": 100
})

Documentation