app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
import torch
import spaces  
import os
IS_SPACES_ZERO = os.environ.get("SPACES_ZERO_GPU", "0") == "1"
IS_SPACE = os.environ.get("SPACE_ID", None) is not None

device = "cuda" if torch.cuda.is_available() else "cpu"
#device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
#dtype = torch.float16
LOW_MEMORY = os.getenv("LOW_MEMORY", "0") == "1"
print(f"Using device: {device}")
#print(f"Using dtype: {dtype}")
print(f"low memory: {LOW_MEMORY}")
model_name = "ruslanmv/Medical-Llama3-8B"
# Move model and tokenizer to the CUDA device
model = AutoModelForCausalLM.from_pretrained(model_name).to(device)
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token
@spaces.GPU
def askme(symptoms, question):
    sys_message = '''\
    You are an AI Medical Assistant trained on a vast dataset of health information. Please be thorough and
    provide an informative answer. If you don't know the answer to a specific medical inquiry, advise seeking professional help.
    '''
    content = symptoms + " " + question
    messages = [{"role": "system", "content": sys_message}, {"role": "user", "content": content}]
    prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
   # inputs = tokenizer(prompt, return_tensors="pt").to(device)  # Ensure inputs are on CUDA device
   # outputs = model.generate(**inputs, max_new_tokens=200, use_cache=True)
   # response_text = tokenizer.batch_decode(outputs)[0].strip()



    # Tokenize all prompts and batch them
    tokenized_inputs = tokenizer(prompts, return_tensors="pt", padding=True, truncation=True).to(device)
    outputs = model.generate(**tokenized_inputs, max_new_tokens=200, use_cache=True)

    # Decode responses
    response_texts = tokenizer.batch_decode(outputs, skip_special_tokens=True)

    
    
    # Remove system messages and content
    #response_text = response_text.replace(sys_message, "").replace(content, "").strip()
    # Extract only the assistant's response
    #assistant_response = response_text.split("<|im_start|>assistant")[1].strip().replace('<|im_end', '')
    # Extract only the assistant's response
    
    
    
    
       # Extract assistant's responses
    assistant_responses = []
    for response_text in response_texts:
        assistant_response = response_text.split("assistant")[1].strip().replace('<|im_end', '')
        assistant_responses.append(assistant_response)

    return assistant_responses
    
    
    
    
    
    
    
    
    
    return assistant_response
# Example usage
symptoms = '''\
I'm a 35-year-old male and for the past few months, I've been experiencing fatigue,
increased sensitivity to cold, and dry, itchy skin.
'''
question = '''\
Could these symptoms be related to hypothyroidism?
If so, what steps should I take to get a proper diagnosis and discuss treatment options?
'''

examples = [
    [symptoms, question]
]

iface = gr.Interface(
    fn=askme,
    inputs=["text", "text"],
    outputs="text",
    examples=examples,
    title="Medical AI Chatbot",
    description="Ask me a medical question!"
)

iface.launch()