app.py

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModelWithLMHead
import gradio
import torch

username = "SrdharMahendran"      # change it to your HuggingFace username
model_path = username + '/MedQnA_Model'
tokenizer_path = username + '/MedQnA_Tokenizer'

loaded_model = AutoModelWithLMHead.from_pretrained(model_path)
loaded_tokenizer = AutoTokenizer.from_pretrained(tokenizer_path)

def generate_query_response(prompt, max_length=200):

    model = loaded_model
    tokenizer = loaded_tokenizer
    input_ids = tokenizer.encode(prompt, return_tensors="pt")      # 'pt' for returning pytorch tensor

    # Create the attention mask and pad token id
    attention_mask = torch.ones_like(input_ids)
    pad_token_id = tokenizer.eos_token_id

    output = model.generate(
        input_ids,
        max_length=max_length,
        num_return_sequences=1,
        attention_mask=attention_mask,
        pad_token_id=pad_token_id
    )

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

iface = gradio.Interface(fn=generate_query_response, 
                     inputs="text", 
                     outputs="text",
                     title = "MedQnA Application")
iface.launch()