app.py

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch
import gradio as gr

# Load model and tokenizer
model_name = "Amitabhdas/code-summarizer-python"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Define summarization function
def summarize_code_with_attention(code_snippet):
    inputs = tokenizer(
        code_snippet,
        max_length=512,
        truncation=True,
        padding="max_length",
        return_tensors="pt"
    )
    with torch.no_grad():
        outputs = model.generate(
            inputs.input_ids,
            max_length=150,
            num_beams=4,
            early_stopping=True,
            output_attentions=True,
            return_dict_in_generate=True
        )
    summary = tokenizer.decode(outputs.sequences[0], skip_special_tokens=True)
    attention = str(outputs.encoder_attentions)  # Convert to string for safe JSON
    input_shape = list(inputs.input_ids.shape)
    
    return {
        "summary": summary,
        "attention": attention,
        "input_shape": input_shape
    }

# Launch the Gradio interface
demo = gr.Interface(
    fn=summarize_code_with_attention,
    inputs=gr.Textbox(lines=10, label="Enter Python code"),
    outputs="json",
    title="Code Summarizer API"
)

demo.launch(share=True)