app.py

# first time - ran this on command line first in directory where I am putting app.py: git clone https://huggingface.co./spaces/brdemorin/chat
# this will create a "chat" directory. This "app.py" file will need to be saved to that chat directory
# change directory in command line to: C:\Users\brian.morin\Documents\HuggingFace\chat
# then do the below. Must do the below everytime I make changes to app.py
# 1 -> change directory to: cd C:\Users\brian.morin\Documents\HuggingFace\chat
# 2 git add app.py
# 3. git commit -m "Add application file"
# 4. git push
# 5. # I'm not sure if I actually need to do this: in your terminal, navigate to the directory containing your app.py file and run the command: streamlit run app.py
# 6. # then navigate here: https://huggingface.co./spaces/brdemorin/chat

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer

x = st.slider('Select a value')
st.write(x, 'squared is', x * x)

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("brdemorin/Phi3-custom_60-steps")
model = AutoModelForCausalLM.from_pretrained("brdemorin/Phi3-custom_60-steps")

# Create a text input for the user to enter their message
user_input = st.text_input("Enter your message:")

# When the user enters a message and presses enter, generate a response
if user_input:
    # Encode the user's message and pass it to the model
    input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors='pt')
    generated_response_ids = model.generate(input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)

    # Decode the model's output IDs to a string and display it
    generated_response = tokenizer.decode(generated_response_ids[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
    st.write(generated_response)