app.py

from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
import torch
import pickle
import streamlit as st
from huggingface_hub import InferenceClient

client = InferenceClient(
    "mistralai/Mistral-7B-Instruct-v0.1"
)


device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")

from translate import Translator

def init_session_state():
    if 'history' not in st.session_state:
        st.session_state.history = ""


temperature=0.9
max_new_tokens=256
top_p=0.95
repetition_penalty=1.0


generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=42,
    )

def format_prompt(message, history):
  prompt = "<s>"
  for user_prompt, bot_response in history:
    prompt += f"[INST] {user_prompt} [/INST]"
    prompt += f" {bot_response}</s> "
  prompt += f"[INST] {message} [/INST]"
  return prompt


# Initialize session state
init_session_state()

# pipe = pipeline("text2text-generation", model="google/flan-t5-base")
pipe = pipeline("text-generation", model="microsoft/Phi-3-mini-4k-instruct", trust_remote_code=True)
# pipe = pipeline("text-generation", model="GeneZC/MiniChat-1.5-3B")
# pipe = pipeline("text-generation", model="mistralai/Mistral-7B-Instruct-v0.2")
# model_name = "MoritzLaurer/mDeBERTa-v3-base-mnli-xnli"
# tokenizer = AutoTokenizer.from_pretrained(model_name)
# model = AutoModelForSequenceClassification.from_pretrained(model_name)

classifier = pipeline("zero-shot-classification", model="MoritzLaurer/mDeBERTa-v3-base-mnli-xnli")

# with open('chapter_titles.pkl', 'rb') as file:
#     titles_astiko = pickle.load(file)
# labels1 = ["κληρονομικό", "ακίνητα", "διαζύγιο"]
# # labels2 = ["αποδοχή κληρονομιάς", "αποποίηση", "διαθήκη"]
# # labels3 = ["μίσθωση", "κυριότητα", "έξωση", "απλήρωτα νοίκια"]


# titles_astiko = ["γάμος", "αλλοδαπός", "φορολογία", "κληρονομικά", "στέγη", "οικογενειακό", "εμπορικό","κλοπή","απάτη"]
# Load dictionary from the file using pickle
with open('my_dict.pickle', 'rb') as file:
    dictionary = pickle.load(file)

def classify(text,labels):
    output = classifier(text, labels, multi_label=False)
    
    return output


text = st.text_input('Enter some text:')  # Input field for new text

if text:

    labels = list(dictionary)
    
    output = classify(text,labels)

    output = output["labels"][0]

    labels = list(dictionary[output])

    output2 = classify(text,labels)

    output2 = output2["labels"][0]


    answer = dictionary[output][output2]

    # Create a translator object with specified source and target languages
    translator = Translator(from_lang='el', to_lang='en')
    translator2 = Translator(from_lang='en', to_lang='el')

    st.text("H ερώτηση σας σχετίζεται με " + output+ " δίκαιο")
 

# Translate the text from Greek to English
    answer = translator.translate(answer)
    text = translator.translate(text)

    st.text("Πιο συγκεκριμένα σχετίζεται με " + output2)


# text_to_translate2 = text[499:999]
# translated_text2 = translator.translate(text_to_translate2)

    

    # st.session_state.history += "Based on this info only:" + answer +" ,answer this question, by reasoning step by step:" + text  # Add new text to history
    # out = pipe(st.session_state.history, max_new_tokens=256)  # Generate output based on history

    history = st.session_state.history
    prompt =  "Based on this info only:" + answer +" ,answer this question, by reasoning step by step:" + text
    formatted_prompt = format_prompt(prompt, history)

    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""

    for response in stream:
        output += response.token.text
        yield output
    return output


    # st.text(st.session_state.history)
    
    # translated_text2 = translator2.translate(out[0]['generated_text'])
    translated_text2 = translator2.translate(output)

    st.text(translated_text2)

    # with st.expander("View Full Output", expanded=False):
    #     st.write(translated_text2, allow_output_mutation=True)

    # st.text(translated_text2)
    # st.text("History: " + st.session_state.history)

    # st.text(output)
    # st.text(output2)

    # st.text(answer)