SUBLLM: A Novel Efficient Architecture with Token Sequence Subsampling for LLM

Open in new window