README.md

---
language:
- en
- ro
tags:
- translation
license: cc-by-4.0
datasets:
- quickmt/quickmt-train.ro-en
model-index:
- name: quickmt-ro-en
  results:
  - task:
      name: Translation ron-eng
      type: translation
      args: ron-eng
    dataset:
      name: flores101-devtest
      type: flores_101
      args: ron_Latn eng_Latn devtest
    metrics:
    - name: BLEU
      type: bleu
      value: 44.93
    - name: CHRF
      type: chrf
      value: 69.3
    - name: COMET
      type: comet
      value: 89.31
---


# `quickmt-ro-en` Neural Machine Translation Model 

`quickmt-ro-en` is a reasonably fast and reasonably accurate neural machine translation model for translation from `ro` into `en`.


## Try it on our Huggingface Space

Give it a try before downloading here: https://huggingface.co/spaces/quickmt/QuickMT-Demo


## Model Information

* Trained using [`eole`](https://github.com/eole-nlp/eole) 
* 200M parameter transformer 'big' with 8 encoder layers and 2 decoder layers
* 32k separate Sentencepiece vocabs
* Expested for fast inference to [CTranslate2](https://github.com/OpenNMT/CTranslate2) format
* Training data: https://huggingface.co/datasets/quickmt/quickmt-train.ro-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:

```bash
git clone https://github.com/quickmt/quickmt.git
pip install ./quickmt/

quickmt-model-download quickmt/quickmt-ro-en ./quickmt-ro-en
```

Finally use the model in python:

```python
from quickmt import Translator

# Auto-detects GPU, set to "cpu" to force CPU inference
t = Translator("./quickmt-ro-en/", device="auto")

# Translate - set beam size to 1 for faster speed (but lower quality)
sample_text = 'Dr. Ehud Ur, profesor de medicină la Universitatea Dalhousie din Halifax, Nova Scotia, și președinte al diviziei clinice și științifice a Asociației Canadiene pentru Diabet, a atras atenția că cercetarea se află încă în fază incipientă.'

t(sample_text, beam_size=5)
```

> 'Dr Ehud Ur, a professor of medicine at Dalhousie University in Halifax, Nova Scotia, and president of the clinical and scientific division of the Canadian Diabetes Association, said the research was still in its infancy.'

```python
# 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)
```

> 'Dr Ehud Ur, MD, clinical clinical and scientific chair of the Canadian Diabetes Association’s Scientific and Clinical Division, stated: “the study is still in its infancy.'

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](https://libretranslate.com/) which also uses `ctranslate2` and `sentencepiece`. A model in safetensors format to be used with `eole` is also provided.


## Metrics

`bleu` and `chrf2` are calculated with [sacrebleu](https://github.com/mjpost/sacrebleu) on the [Flores200 `devtest` test set](https://huggingface.co/datasets/facebook/flores) ("ron_Latn"->"eng_Latn"). `comet22` with the [`comet`](https://github.com/Unbabel/COMET) library and the [default model](https://huggingface.co/Unbabel/wmt22-comet-da). "Time (s)" is the time in seconds to translate the flores-devtest dataset (1012 sentences) on an RTX 4070s GPU with batch size 32 (faster speed is possible using a larger batch size).


|                                  |   bleu |   chrf2 |   comet22 |   Time (s) |
|:---------------------------------|-------:|--------:|----------:|-----------:|
| quickmt/quickmt-ro-en            |  44.93 |   69.3  |     89.31 |       1.37 |
| facebook/nllb-200-distilled-600M |  43.3  |   68.16 |     88.86 |      22.01 |
| facebook/nllb-200-distilled-1.3B |  45.32 |   69.63 |     89.47 |      38.38 |
| facebook/m2m100_418M             |  34.44 |   62.75 |     85.64 |      18.57 |
| facebook/m2m100_1.2B             |  40.96 |   66.93 |     88.41 |      35.87 |