in src/open-r1-multimodal/local_scripts/create_vision_cot_data.py [0:0]
def main():
dataset_path = "path/to/your/dataset"
full_dataset = load_from_disk(dataset_path)
processed_dataset = full_dataset.map(
function=partial(process_single_item),
num_proc=256,
desc="Processing dataset with GPT-4o",
keep_in_memory=True,
)
output_path = f"{dataset_path}_processed"
processed_dataset.save_to_disk(output_path)
print(f"Processed dataset saved to: {output_path}")