app.py

import torch 
from transformers import AutoTokenizer, AutoModelForCausalLM
import gradio as gr 
import spaces 
device = 'cuda' if torch.cuda.is_available() else 'cpu'

@spaces.GPU
def load_model():
    tokenizer = AutoTokenizer.from_pretrained("alibidaran/medical_transcription_generator")
    model = AutoModelForCausalLM.from_pretrained("alibidaran/medical_transcription_generator").to(device)
    return model,tokenizer
@spaces.GPU
def generate_text(Text,Max_length,Temperature):
  model,tokenizer=load_model()
  torch.manual_seed(32)
  tokenizer.pad_token_id=tokenizer.eos_token_id
  with torch.no_grad():
    input_ids = tokenizer(Text, return_tensors="pt")["input_ids"].to(device)
    attn_mask=tokenizer(Text, return_tensors="pt")["attention_mask"].to(device)
    output=model.generate(input_ids=input_ids,attention_mask=attn_mask,max_new_tokens=Max_length,do_sample=True, temperature=Temperature, top_p=0.90,top_k=10)
    return tokenizer.decode(output[0])
demo=gr.Interface(
    generate_text,
    ['text',
     gr.Slider(50,2000,value=100,step=10),
     gr.Slider(0,2,value=0.7,step=0.1)],
    'text',
      theme=gr.themes.Base(primary_hue='blue',secondary_hue='cyan'),
      description="Medical Trasncript Generator"
)
demo.launch()