https://github.com/thisisnotjustin/deng
Transformer Architecture for translation between English and Dutch
https://github.com/thisisnotjustin/deng
machine-learning python pytorch
Last synced: about 1 month ago
JSON representation
Transformer Architecture for translation between English and Dutch
- Host: GitHub
- URL: https://github.com/thisisnotjustin/deng
- Owner: ThisIsNotJustin
- Created: 2025-03-07T05:45:58.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-03-07T08:52:47.000Z (over 1 year ago)
- Last Synced: 2025-03-14T23:46:22.497Z (over 1 year ago)
- Topics: machine-learning, python, pytorch
- Language: Python
- Homepage:
- Size: 4.88 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Deng (Dutch - English)
PyTorch implementation of a Transformer model for translation between English and Dutch.
## Features
- Bidirectional translation
- Transformer Architecture
- Tokenization
- Early Stop and Model Checkpoints
- CUDA support
## Project Structure
```
deng/
│
├── deng.py # Main training module
├── test.py # Testing and inference module
├── eng-dutch.tsv # Training data
└── checkpoint/ # Saved model checkpoints
```
## Usage
### Training
To train the model:
```bash
python deng.py
```
The model will be saved automatically in the `checkpoint` directory when the validation loss improves.
### Testing
To test the trained model:
```bash
python test.py
```
## Dataset
The model is trained on the Tatoeba English-Dutch dataset stored in TSV format with the following columns:
- eng_id: English sentence ID
- eng: English sentence
- d_id: Dutch sentence ID
- dutch: Dutch sentence