app.py

import os 
import gradio as grad
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, TextStreamer

auth_token = os.environ.get("auth_token")
model_c = 'nadsoft/faseeh.v.9'
tokenizer = AutoTokenizer.from_pretrained(model_c,src_lang='ar_AR', tgt_lang='en_XX',use_auth_token=auth_token)
model = AutoModelForSeq2SeqLM.from_pretrained(model_c,use_auth_token=auth_token)
streamer = TextStreamer(tokenizer)

def translate(text):
  inputs = tokenizer(text, return_tensors="pt")
  outputs = model.generate(**inputs, max_length=128, num_beams=1,streamer=streamer)
  response = tokenizer.decode(outputs[0], skip_special_tokens=True)

  return response
grad.Interface(translate, inputs=["text"], outputs=["text"]).launch()