File size: 4,668 Bytes
0d3e097 bf1a04b 0d3e097 28dc7cd 0d3e097 28dc7cd 0d3e097 28dc7cd 0d3e097 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 | ---
language:
- en
- it
tags:
- translation
license: cc-by-4.0
datasets:
- quickmt/quickmt-train.it-en
model-index:
- name: quickmt-en-it
results:
- task:
name: Translation eng-ita
type: translation
args: eng-ita
dataset:
name: flores101-devtest
type: flores_101
args: eng_Latn ita_Latn devtest
metrics:
- name: BLEU
type: bleu
value: 30.52
- name: CHRF
type: chrf
value: 59.72
- name: COMET
type: comet
value: 87.55
---
<a href="https://huggingface.co/spaces/quickmt/quickmt-gui"><img src="https://huggingface.co/datasets/huggingface/badges/resolve/main/open-in-hf-spaces-lg-dark.svg" alt="Open in Spaces"></a>
# `quickmt-en-it` Neural Machine Translation Model
`quickmt-en-it` is a reasonably fast and reasonably accurate neural machine translation model for translation from `en` into `it`.
## Model Information
* Trained using [`eole`](https://github.com/eole-nlp/eole)
* 185M parameter transformer 'big' with 8 encoder layers and 2 decoder layers
* 20k sentencepiece vocabularies
* Exported for fast inference to [CTranslate2](https://github.com/OpenNMT/CTranslate2) format
* Training data: https://huggingface.co/datasets/quickmt/quickmt-train.it-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](github.com/quickmt/quickmt).
```bash
git clone https://github.com/quickmt/quickmt.git
pip install ./quickmt/
```
Finally, use the model in python:
```python
from quickmt import Translator
from huggingface_hub import snapshot_download
# Download Model (if not downloaded already) and return path to local model
# Device is either 'auto', 'cpu' or 'cuda'
t = Translator(
snapshot_download("quickmt/quickmt-en-it", ignore_patterns="eole-model/*"),
device="cpu"
)
# 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)
> 'Il dottor Ehud Ur, professore di medicina alla Dalhousie University di Halifax, in Nuova Scozia e presidente della divisione clinica e scientifica della Canadian Diabetes Association, ha avvertito che la ricerca è ancora agli inizi.'
# 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)
> "Il Dott. Ehud Ur, professore di medicina alla Dalhousie University di Halifax, Nuova Scozia e presidente del dipartimento scientifico e clinica dell'Associazione Canada-del diabete ha avvertito che questa ricerca si trova ancora in pieno giorno."
```
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`.
## 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) ("eng_Latn"->"ita_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 (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).
| | bleu | chrf2 | comet22 | Time (s) |
|:---------------------------------|-------:|--------:|----------:|-----------:|
| quickmt/quickmt-en-it | 30.52 | 59.72 | 87.55 | 1.06 |
| Helsink-NLP/opus-mt-en-it | 27.58 | 57.66 | 85.1 | 4.42 |
| facebook/nllb-200-distilled-600M | 28.58 | 58.3 | 87.3 | 24.93 |
| facebook/nllb-200-distilled-1.3B | 30.92 | 59.89 | 88.52 | 43.8 |
| facebook/m2m100_418M | 23.81 | 54.34 | 82.64 | 20.34 |
| facebook/m2m100_1.2B | 27.22 | 57.19 | 86.23 | 39.98 |
`quickmt-en-it` is the fastest and second highest quality next to `nllb-200-distilled-1.3B`.
|