app.py

import streamlit as st
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load the fine-tuned model and tokenizer
model_repo_path = 'sabssag/Latex_to_Python_T5-small'
model = AutoModelForSeq2SeqLM.from_pretrained(model_repo_path)
tokenizer = AutoTokenizer.from_pretrained(model_repo_path)

# Function to generate Python code from LaTeX expression
def generate_code(latex_expr, tokenizer, model, max_length=512):
    # Tokenize the input LaTeX expression
    inputs = tokenizer.encode(
        latex_expr,
        return_tensors='pt',
        max_length=max_length,
        truncation=True
    )

    # Generate the output
    with torch.no_grad():
        outputs = model.generate(
            inputs,
            max_length=max_length,
            num_beams=5,  # Number of beams for beam search
            early_stopping=True
        )

    # Decode the output tokens to string
    generated_code = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return generated_code

# Streamlit app layout
st.title("LaTeX to Python Code Generator")

# Define session state keys
if 'latex_expr' not in st.session_state:
    st.session_state.latex_expr = ""

# User input for LaTeX expression
latex_input = st.text_area("Enter the LaTeX Expression", value=st.session_state.latex_expr, height=150)

# Update session state with the new LaTeX expression
if st.button("Generate Code"):
    if latex_input:
        st.session_state.latex_expr = latex_input
        with st.spinner("Generating Python Code..."):
            try:
                generated_code = generate_code(latex_expr=st.session_state.latex_expr, tokenizer=tokenizer, model=model)
                # Display the generated code
                st.subheader("Generated Python Code")
                st.code(generated_code, language='python')
            except Exception as e:
                st.error(f"Error during code generation: {e}")
    else:
        st.warning("Please enter a LaTeX expression to generate Python code.")