backend/german_polish.py

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch
from backend.logger.logger import log_to_gsheet
import threading

model_id = "DebasishDhal99/polish-to-german-toponym-model-opus-mt-pl-de"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")


def polish_to_german_translation(polish_place):
    inputs = tokenizer(polish_place, return_tensors="pt", padding=True, truncation=True)
    inputs = {k: v.to(device) for k, v in inputs.items()}


    with torch.no_grad():
        outputs = model.generate(**inputs, max_length=50)
    
    german_place = tokenizer.decode(outputs[0], skip_special_tokens=True)
    threading.Thread(
        target=log_to_gsheet,
        args=(polish_place, german_place, "Polish", "German"),
        daemon=True,
    ).start()
    # log_to_gsheet(polish_place, german_place, "Polish", "German")
    return german_place