Code for the EMNLP 2021 Paper "AfroMT: Pretraining Strategies and Reproducible Benchmarks for Translation of 8 African Languages" by Machel Reid, Junjie Hu, Graham Neubig, Yutaka Matsuo
# AfroMT
Code for the EMNLP 2021 Paper AfroMT: Pretraining Strategies and Reproducible Benchmarks for Translation of 8 African Languages.
```bibtex
@inproceedings{reid21afromt,
title = {Afro{MT}: Pretraining Strategies and Reproducible Benchmarks for Translation of 8 African Languages},
author = {Machel Reid and Junjie Hu and Graham Neubig and Yutaka Matsuo},
booktitle = {Conference on Empirical Methods in Natural Language Processing (EMNLP)},
address = {Punta Cana, Dominican Republic},
month = {November},
url = {
arxiv.org,
year = {2021}
}
```
## Clone the repo
```bash
git clone git@github.com:machelreid/afromt
```
## Data
Please use `gdown` to download the data
```bash
cd afromt
gdown --id 1Qj3IXQ9kusaeHtGYRVPzMnGm_dQvvfJP
tar -xf afromt.tar.xz
```
## Installation
Run the following commands
```bash
cd training
# install apex for fp16 for faster training
git clone
github.com
cd apex
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
cd ../
# install fairseq
pip install -e .
cd ../
```
## Model usage
Download the model as follows:
```bash
gdown --id 1A9kbWHnMrjFwgq9x8rpnU4aATx_Y4b3b
tar -xf afrobart.tar.xz # and you will get afrobart.pt which is the model file
```
### Preprocessing
First be sure to install sentencepiece
```bash
TGT_LANG=#whatever target language you choose
cd afromt # the data folder
cd en-$TGT_LANG
bash ../../preprocess.sh $TGT_LANG ../../
cd ../../
```
to preprocess \$TGT_LANG (e.g. `zu` for Zulu, or `xh` for Xhosa). This command will produce a `data-bin-afromt` folder in the folder for the en-\$TGT_LANG language pair.
### Training
More details within the script itself if you want to tweak training arguments!
```bash
bash train_afrobart.sh afromt/en-$TGT_LANG/data-bin-afromt afromt/en-$TGT_LANG/model_output $TGT_LANG
```
and you should see your model training!
### Evaluation
Once training is over, you can generate from the model as …