app.py

from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, pipeline
import gradio as gr
import torch
import spaces

model_id = "deepapaikar/Llama_SCplusQA_10epochs"
print("Before loading model")
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16,
    bnb_4bit_use_double_quant=True,
)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    return_dict=True,
    device_map="auto",
    trust_remote_code=True,
    config=bnb_config,
)

tokenizer = AutoTokenizer.from_pretrained(model_id)
tokenizer.pad_token = tokenizer.eos_token

pipe = pipeline(
    task="text-generation",
    model=model,
    tokenizer=tokenizer,
    max_length=100,
    trust_remote_code=True,
)
print("Model loaded successfully")
#system_message = "Answer the questions truthfully and to the point."

def generate_response(query, history):
    ans = pipe(f"[INST] {query} [/INST]")
    result= ans[0]['generated_text']
    return result

demo = gr.ChatInterface(generate_response)

if __name__ == "__main__":
    demo.launch(share=True)