import gradio as gr from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM # Available models for code explanation MODELS = { "CodeT5-small-sum": "Salesforce/codet5-small", "CodeT5-base-sum": "Salesforce/codet5-base-multi-sum", "CodeBERTa": "huggingface/CodeBERTa-small-v1" } def load_model(model_name): tokenizer = AutoTokenizer.from_pretrained(model_name) model = AutoModelForSeq2SeqLM.from_pretrained(model_name) # Use summarization for CodeT5, text2text-generation for others if "codet5" in model_name.lower(): return pipeline("summarization", model=model, tokenizer=tokenizer) else: return pipeline("text2text-generation", model=model, tokenizer=tokenizer) def explain_code(code, model_choice): model_id = MODELS[model_choice] explainer = load_model(model_id) lines = code.splitlines() line_explanations = [] for idx, line in enumerate(lines): if not line.strip(): continue # Skip empty lines input_length = len(line.split()) max_length = max(1, input_length // 2) gen_kwargs = dict(max_length=max_length, do_sample=False) if "codet5" in model_id.lower(): result = explainer(line, **gen_kwargs)[0].get('summary_text', '') else: prompt = f"Explain this line of code:\n{line}" result = explainer(prompt, **gen_kwargs)[0].get('generated_text', '') line_explanations.append(f"Line {idx+1}: {line}\nExplanation: {result.strip()}\n") return "\n".join(line_explanations) def main(): with gr.Blocks() as demo: gr.Markdown("# CodeExplainerBot (Hugging Face UI)") model_choice = gr.Dropdown(list(MODELS.keys()), label="Select a model", value="CodeT5-small-sum") code_input = gr.Textbox(label="Paste your code here", lines=20) output = gr.Textbox(label="Line-by-line Explanation", lines=20) explain_btn = gr.Button("Explain Code") explain_btn.click(fn=explain_code, inputs=[code_input, model_choice], outputs=output) demo.launch() if __name__ == "__main__": main()