|
--- |
|
language: african-languages |
|
tags: |
|
- african-languages |
|
- machine-translation |
|
- text |
|
license: apache-2.0 |
|
model-index: |
|
- name: Maakhane Benchmark ModelsFon XLSR Wav2Vec2 Large 53 |
|
results: |
|
- task: |
|
name: Machine Translation |
|
type: machine-translation |
|
dataset: |
|
name: masakhane benchmarks |
|
args: african-languages |
|
|
|
--- |
|
# Masakhane Benchmark Models |
|
|
|
I created this demo for very easy interaction with the [benchmark models on Masakhane](https://github.com/masakhane-io/masakhane-mt/tree/master/benchmarks) which were trained with [JoeyNMT](https://github.com/chrisemezue/joeynmt)(my forked version). |
|
|
|
To access the space click [here](https://huggingface.co/spaces/chrisjay/masakhane-benchmarks). |
|
|
|
To include your language, all you need to do is: |
|
1. Create a folder in the format *src-tgt/main* for your language pair, if it does not exist. |
|
2. Inside the *main* folder put the following files: |
|
i. model checkpoint. Rename it to `best.ckpt`. |
|
ii. `config.yaml` file. This is the JoeyNMT config file which loads the model an pre-processing parameters. |
|
iii. `src_vocab.txt` file. |
|
iv. `trg_vocab.txt` file. |
