========== ORIGINAL PROMPT ==========
def build_open_ended_prompt(sample):
    """Build an open-ended prompt (no MCQ options)."""
    desc = sample.get('description', '')
    question = sample.get('question', '')

    prompt = f"""Look at the image and answer the physics question.

{desc}

{question}


========== ORIGINAL GENERATE ==========
                output_ids = model.generate(**inputs, max_new_tokens=MAX_NEW_TOKENS)

            generated = output_ids[0][inputs.input_ids.shape[1]:]
            response = processor.decode(generated, skip_special_tokens=True)
        except Exception as e:
            response = f"ERROR: {str(e)}"

========== MY PROMPT ==========
def build_open_ended_prompt(sample):
    q = sample.get("question", "")
    return f"Look at this image and answer the physics question. Think step by step and put your final answer in \\boxed{{}}.\n\nQuestion: {q}"

def run_inference_on_gpu(gpu_id, model_path, samples, output_file):
    """Run inference for a subset of samples on a specific GPU."""

========== MY GENERATE ==========
                output_ids = model.generate(**inputs, max_new_tokens=2048, temperature=0.1, do_sample=False)
            
            output_text = processor.batch_decode(output_ids[:, inputs.input_ids.shape[1]:], 
                                                  skip_special_tokens=True)[0]