File size: 2,259 Bytes
df26ca1 ceab878 | 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 | ---
license: apache-2.0
datasets:
- Helsinki-NLP/tatoeba
- openlanguagedata/flores_plus
- facebook/bouquet
language:
- en
- it
metrics:
- bleu
- comet
- chrf
pipeline_tag: translation
---
# OPUS-MT-tiny-eng-ita
Distilled model from a Tatoeba-MT Teacher: [Tatoeba-MT-models/deu+eng+fra+por+spa-itc/opusTCv20230926max50+bt+jhubc_transformer-big_2024-05-30.](https://object.pouta.csc.fi/Tatoeba-MT-models/deu+eng+fra+por+spa-itc/opusTCv20230926max50+bt+jhubc_transformer-big_2024-05-30.zip), which has been trained on the [Tatoeba](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/data) dataset.
We used the [OpusDistillery](https://github.com/Helsinki-NLP/OpusDistillery) to train new a new student with the tiny architecture, with a regular transformer decoder.
For training data, we used [Tatoeba](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/data).
The configuration file fed into OpusDistillery can be found [here](https://github.com/Helsinki-NLP/OpusDistillery/blob/main/configs/opustranslate_hf/config.op.en-it.yml).
## How to run
```python
from transformers import MarianMTModel, MarianTokenizer
model_name = "Helsinki-NLP/opus-mt_tiny_eng-ita"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
tok = tokenizer("The area is also home to species of animals and birds with a wide variety.", return_tensors="pt").input_ids
output = model.generate(tok)[0]
tokenizer.decode(output, skip_special_tokens=True)
```
## Benchmarks
### Teacher
| testset | BLEU | chr-F | COMET|
|-----------------------|-------|-------|-------|
| Flores+ | 30.6 | 59.4 | 0.8600 |
| Bouquet | 55.0 | 74.9 | 0.8865 |
### Student
| testset | BLEU | chr-F | COMET |
|-----------------------|-------|-------|-------|
| Flores+ | 27.6 | 57.5 | 0.8410 |
| Bouquet | 49.8 | 71.4 | 0.8785 |
## Marian models
We also provide Marian-compatible versions of this model. To use them, compile [Marian](https://marian-nmt.github.io/quickstart/) and run decoding with `marian-decoder`, for example:
```bash
marian-decoder \
-i input.txt \
-c final.model.npz.best-perplexity.npz.decoder.yml \
-m final.model.npz.best-perplexity.npz \
-v vocab.spm vocab.spm
|