reasoning-machines / cocogen Goto Github PK

Hi,
In the code below from the script src/api/query_openai_over_tasks.py :

if task['input_prompt_code'] in cache:
        logging.info(
            f"Task {task_idx} > Using cached result for {task['input_prompt_code']}")
        codex_response = cache[task['input_prompt_code']]["codex_response"]
    else:
        codex_response = query_codex(task, prompt_text, engine, max_tokens=max_tokens)


    completed_code = get_completed_code(task, codex_response)

    graph = converter.python_to_graph(completed_code)

it seems the converter is applied only after the response is generated from the openai. Can you please verify whether this is correct? It seems the prompts are not getting changed when different prompting schemes are applied before calling openai.

Hi @madaan , Thanks for the great work!

I wonder which code structure you used for proscript to get the best results?

According to fig1(b) is the Node class with children attributes whereas in the repo, I see a different format. Appreciate your comments on this :)