trinity.common.models.mm_utils module# trinity.common.models.mm_utils.build_multi_modal_inputs(prompt: str, raw_mm_data: Dict[str, Any], processor: Any, **kwargs) → Dict[str, Any][source]# Preprocess multi-modal data and build multi-modal inputs Adapted from: verl/utils/dataset/rl_dataset.py trinity.common.models.mm_utils.attach_images_to_messages(messages, raw_mm_data)[source]#
