espnet2.sds.end_to_end.mini_omni.litgpt.model.build_mask_cache

source

espnet2.sds.end_to_end.mini_omni.litgpt.model.build_mask_cache(max_seq_length: int, device: device | None = None) → Tensor