Building Omage — AI-First Python Library. Making PyTorch feel like pseudocode. Open source enthusiast.

Omage: Build AI Models in 5 Lines of Python

The Problem

PyTorch is powerful but verbose. Training a simple classifier takes 50+ lines of boilerplate code.

The Solution

I built Omage — a high-level Python library that abstracts PyTorch into readable, concise code.

Before (PyTorch)

python
import torch
import torch.nn as nn
import torch.optim as optim

class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.layers = nn.Sequential(
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Dropout(0.2),
            nn.Linear(64, 10)
        )

    def forward(self, x):
        return self.layers(x)

model = Model()
optimizer = optim.Adam(model.parameters(), lr=0.001)
criterion = nn.CrossEntropyLoss()

# ... 30 more lines for training loop

##After (Omage)

import omage as og

ai = og.model(
    type="classifier",
    layers=[og.dense(128), og.dropout(0.2), og.dense(10)],
    optimizer=og.adam(lr=0.001),
    loss=og.cross_entropy(),
)
ai.train(data, epochs=10)

##Features

Auto GPU detection — runs on CUDA, MPS, or CPU automatically

Model Zoo — load ResNet, GPT-2, BERT, YOLO in one line

CLI —  omage run ,  omage compile ,  omage zoo list 

##Installation

pip install git+https://github.com/Omage-Python-Library/Omage.git

##Quick Start

import omage as og

data = og.load("data.csv").clean().normalize().split(0.8)

model = og.model(
    type="classifier",
    layers=[og.dense(128), og.dense(10)],
)
model.train(data, epochs=10)
model.save("model.omg")

##Links

GitHub: github.com/Omage-Python-Library/Omage

##Feedback

This is v0.1.1 (alpha). Would love feedback from the community!