Model Card for Model ID

This model is a translator into Lithuanian and vice versa. It was trained on the following datasets:

Model Usage

import torch
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

from transformers import T5Tokenizer, MT5ForConditionalGeneration

tokenizer = T5Tokenizer.from_pretrained('werent4/mt5TranslatorLT')
model = MT5ForConditionalGeneration.from_pretrained("werent4/mt5TranslatorLT")
model.to(device)
def translate(text, model, tokenizer, device, translation_way = "en-lt"):
    translations_ways = {
        "en-lt": "<EN2LT>",
        "lt-en": "<LT2EN>"
    }
    if translation_way not in translations_ways:
        raise ValueError(f"Invalid translation way. Supported ways: {list(translations_ways.keys())}")
    input_text = f"{translations_ways[translation_way]} {text}"
    encoded_input = tokenizer(input_text, return_tensors="pt", padding=True, truncation=True, max_length=128).to(device)
    with torch.no_grad():
        output_tokens = model.generate(
          **encoded_input,
          max_length=128,
          num_beams=5,
          no_repeat_ngram_size=2,
          early_stopping=True
      )

    translated_text = tokenizer.decode(output_tokens[0], skip_special_tokens=True)
    return translated_text

text = "How are you?"
translate(text, model, tokenizer, device)
`Kaip esate?`

text = "I live in Kaunas"
translate(text, model, tokenizer, device)
`AÅ¡ gyvenu Kaunas`

text = "Mano vardas yra Karolis"
translate(text, model, tokenizer, device, translation_way= "lt-en")
`My name is Karolis`

Model Card Authors

werent4
Mykhailo Shtopko

Model Card Contact

[More Information Needed]

Downloads last month
15
Safetensors
Model size
300M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and the model is not deployed on the HF Inference API.

Space using werent4/mt5TranslatorLT 1