TorchLingo¶

A beginner-friendly PyTorch library for learning Neural Machine Translation

What is TorchLingo?¶

TorchLingo is a compact, educational library designed to teach you the fundamentals of Neural Machine Translation (NMT) using PyTorch. Whether you've never touched PyTorch before or you're looking to understand how Google Translate works under the hood, TorchLingo has you covered.

Perfect for Students

TorchLingo was built for university coursework. The code is intentionally simple, readable, and well-documented—no magic, no hidden complexity.

Run in Google Colab

The easiest way to get started is with Google Colab—no installation required on your machine!

%pip install torchlingo

Colab provides free GPU access which speeds up training significantly. Just go to Runtime → Change runtime type → GPU.

✨ Features¶

Easy to Start

Get your first translation model running in under 5 minutes with our quickstart guide.

Quick Start
Learn by Doing

Interactive Jupyter notebooks walk you through every concept step-by-step.

Tutorials
Readable Code

Every function is documented. Every tensor shape is explained. No black boxes.

API Reference
Two Architectures

Learn both classic LSTM and modern Transformer models side by side.

Model Concepts

🚀 Quick Example¶

Here's a taste of what working with TorchLingo looks like:

from torchlingo.config import Config
from torchlingo.data_processing import NMTDataset, create_dataloaders
from torchlingo.models import SimpleTransformer

# 1. Create a configuration
config = Config(
    batch_size=32,
    learning_rate=1e-4,
    d_model=256,
)

# 2. Load your data
train_dataset = NMTDataset("data/train.tsv", config=config)

# 3. Create your model
model = SimpleTransformer(
    src_vocab_size=len(train_dataset.src_vocab),
    tgt_vocab_size=len(train_dataset.tgt_vocab),
    config=config,
)

# 4. Train and translate!
# (See our tutorials for the complete training loop)

🗺️ Where to Go Next¶

If you want to...	Go here
Install TorchLingo	Installation Guide
Build your first model in 5 minutes	Quick Start
Understand the theory behind NMT	What is NMT?
Follow step-by-step notebooks	Tutorials
Look up a specific function	API Reference

📚 Philosophy¶

TorchLingo follows a few key principles:

Explicit is better than implicit — You can trace every tensor operation
Notebooks are first-class — Every concept is executable
Start simple, add complexity later — Basic tokenization → SentencePiece → Multilingual
Docs that teach — Every docstring explains why, not just what