app.py

import streamlit as st
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

@st.cache_resource
def load_model():
    model_name = "Salesforce/codet5-small"
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
    return tokenizer, model

# Load model
tokenizer, model = load_model()

st.title("Code Generator")
st.write("Generate code snippets from natural language prompts using CodeT5!")

prompt = st.text_area("Enter your coding task:", placeholder="Write a Python function to calculate factorial.")
max_length = st.slider("Maximum length of generated code:", 20, 300, 100)

if st.button("Generate Code"):
    if prompt.strip():
        with st.spinner("Generating code..."):
            inputs = tokenizer(prompt, return_tensors="pt", truncation=True, padding=True)
            outputs = model.generate(inputs.input_ids, max_length=max_length, num_beams=5, temperature=0.7, early_stopping=True)

            st.write("### Debugging: Raw Model Output")
            st.json(outputs.tolist())  # Debugging output

            generated_code = tokenizer.decode(outputs[0], skip_special_tokens=True)
            
            st.write("### Generated Code:")
            st.code(generated_code, language="python")
    else:
        st.warning("Please enter a prompt!")