import streamlit as st
from transformers import pipeline

import torch
from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM, AutoTokenizer


peft_model_id = "JackLiuAngel/bloom-7b1-lora-alfred-team-20240730"
config = PeftConfig.from_pretrained(peft_model_id)
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

# Load the Lora model
model = PeftModel.from_pretrained(model, peft_model_id)

# sentiment_pipeline = pipeline("sentiment-analysis")

st.title("Team info finetuned in bigscience/bloom-7b1")
st.write("ask a question about our team:")

user_input = st.text_input("")



if user_input:
    batch = tokenizer(f"“{user_input}” ->: ", return_tensors='pt')

    with torch.cuda.amp.autocast():
        output_tokens = model.generate(**batch, max_new_tokens=50)

    # print('\n\n', tokenizer.decode(output_tokens[0], skip_special_tokens=True))


    result = tokenizer.decode(output_tokens[0], skip_special_tokens=True)


    st.write(f"reply: {result}")