puettmann/Quadrifoglio-mt-it-en

🍀 Quadrifoglio - A small model for Italian -> English translation

Quadrifoglio is an encoder-decoder transformer model for English-Italian text translation based on bigscience/mt0-small. It was trained on the en-it section of Helsinki-NLP/opus-100 and Helsinki-NLP/europarl.

Usage

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load model and tokenizer from checkpoint directory
tokenizer = AutoTokenizer.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")
model = AutoModelForSeq2SeqLM.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")

def generate_response(input_text):
    input_ids = tokenizer("translate Italian to English:" + input_text, return_tensors="pt").input_ids
    output = model.generate(input_ids, max_new_tokens=256)
    return tokenizer.decode(output[0], skip_special_tokens=True)

text_to_translate = "Vorrei una tazza di tè nero, per favore."
response = generate_response(text_to_translate)
print(response)

As this model is trained on translating sentence pairs, it is best to split longer text into individual sentences, ideally using SpaCy. You can then translate the sentences and join the translations at the end like this:

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import spacy
# First, install spaCy and the Italian language model if you haven't already
# !pip install spacy
# !python -m spacy download it_core_news_sm

nlp = spacy.load("it_core_news_sm")

tokenizer = AutoTokenizer.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")
model = AutoModelForSeq2SeqLM.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")

def generate_response(input_text):
    input_ids = tokenizer("translate Italian to English: " + input_text, return_tensors="pt").input_ids
    output = model.generate(input_ids, max_new_tokens=256)
    return tokenizer.decode(output[0], skip_special_tokens=True)

text = "Ciao, come stai? Oggi è una bella giornata. Spero che tu stia bene."
doc = nlp(text)
sentences = [sent.text for sent in doc.sents]

sentence_translations = []
for i, sentence in enumerate(sentences):
    sentence_translation = generate_response(sentence)
    sentence_translations.append(sentence_translation)

full_translation = " ".join(sentence_translations)
print(full_translation)

Evaluation

Done on the Opus 100 test set.

BLEU

	mt0-Quadrifoglio	mt0-small
BLEU Score	0.4015	0.0159
Precision 1	0.6987	0.2350
Precision 2	0.4702	0.0290
Precision 3	0.3443	0.0076
Precision 4	0.2576	0.0013