raphaelmerx/madlad400-10b-ct2

MADLAD-400-10B-MT, converted to cTranslate2

The google/madlad400-10b-mt model, converted to cTranslate2 format.

Usage:

from huggingface_hub import snapshot_download

model_path = snapshot_download("santhosh/madlad400-3b-ct2")
tokenizer = SentencePieceProcessor()
tokenizer.load(f"{model_path}/sentencepiece.model")
translator = ctranslate2.Translator(model_path, device="cuda" if torch.cuda.is_available() else "cpu")

text = "<2pt> I love pizza!"
input_ids = tokenizer.encode(text, out_type=str)
outputs = model.translate_batch(
    [input_ids],
    batch_type="tokens",
    beam_size=2,
    no_repeat_ngram_size=1,
)

tokenizer.decode(results[0].hypotheses[0])
# Eu adoro pizza!