language:
- en
- ja
tags:
- translation
license: cc-by-4.0
datasets:
- quickmt/quickmt-train.ja-en
model-index:
- name: quickmt-en-ja
results:
- task:
name: Translation eng-jpn
type: translation
args: eng-jpn
dataset:
name: flores101-devtest
type: flores_101
args: eng_Latn jpn_Jpan devtest
metrics:
- name: CHRF
type: chrf
value: 42.04
- name: COMET
type: comet
value: 89.08
quickmt-en-ja Neural Machine Translation Model
quickmt-en-ja is a reasonably fast and reasonably accurate neural machine translation model for translation from en into ja.
Model Information
- Trained using
eole - 185M parameter transformer 'big' with 8 encoder layers and 2 decoder layers
- 20k sentencepiece vocabularies
- Exported for fast inference to CTranslate2 format
- Training data: https://huggingface.co/datasets/quickmt/quickmt-train.ar-en/tree/main
See the eole model configuration in this repository for further details and the eole-model for the raw eole (pytorch) model.
Usage with quickmt
You must install the Nvidia cuda toolkit first, if you want to do GPU inference.
Next, install the quickmt python library and download the model:
git clone https://github.com/quickmt/quickmt.git
pip install ./quickmt/
quickmt-model-download quickmt/quickmt-en-ja ./quickmt-en-ja
Finally use the model in python:
from quickmt import Translator
# Auto-detects GPU, set to "cpu" to force CPU inference
t = Translator("./quickmt-en-ja/", device="auto")
# Translate - set beam size to 5 for higher quality (but slower speed)
sample_text = 'Dr. Ehud Ur, professor of medicine at Dalhousie University in Halifax, Nova Scotia and chair of the clinical and scientific division of the Canadian Diabetes Association cautioned that the research is still in its early days.'
t(sample_text, beam_size=5)
> 'ノバスコシア州ハリファックスのダルハウジー大学の医学教授で、カナダ糖尿病協会の臨床および科学部門の議長であるEhud Ur博士は、研究はまだ初期段階にあると警告しました。'
# Get alternative translations by sampling
# You can pass any cTranslate2 `translate_batch` arguments
t([sample_text], sampling_temperature=1.2, beam_size=1, sampling_topk=50, sampling_topp=0.9)
> 'ノバスコシア州ハリファックスのダルハウジー大学教授で、カナダ糖尿病学会の臨床、科学部会長のエフド・ウル博士がこの研究はまだその初期段階にあると警告した。'
The model is in ctranslate2 format, and the tokenizers are sentencepiece, so you can use ctranslate2 directly instead of through quickmt. It is also possible to get this model to work with e.g. LibreTranslate which also uses ctranslate2 and sentencepiece.
Metrics
chrf2 is calculated with sacrebleu on the Flores200 devtest test set ("eng_Latn"->"jpn_Jpan"). comet22 with the comet library and the default model. "Time (s)" is the time in seconds to translate (using ctranslate2) the flores-devtest dataset (1012 sentences) on an RTX 4070s GPU with batch size 32 (faster speed is possible using a large batch size).
| chrf2 | comet22 | Time (s) | |
|---|---|---|---|
| quickmt/quickmt-en-ja | 42.04 | 89.08 | 1.38 |
| Helsink-NLP/opus-mt-en-ja | 6.41 | 62.91 | 7.09 |
| facebook/nllb-200-distilled-600M | 30.00 | 86.64 | 26.05 |
| facebook/nllb-200-distilled-1.3B | 32.38 | 88.02 | 46.04 |
| facebook/m2m100_418M | 32.73 | 85.09 | 23.29 |
| facebook/m2m100_1.2B | 35.83 | 87.78 | 43.89 |
quickmt-en-ja is the fastest and highest quality by a fair margin.