{"url":"/method/layerdrop","slug":"layerdrop","name":"LayerDrop","full_name":"LayerDrop","full_name_withheld":false,"description_markdown":"**LayerDrop** is a form of structured [dropout](https://paperswithcode.com/method/dropout) for [Transformer](https://paperswithcode.com/method/transformer) models which has a regularization effect during training and allows for efficient pruning at inference time. It randomly drops layers from the Transformer according to an \"every other\" strategy where pruning with a rate $p$ means dropping the layers at depth $d$ such that $d = 0\\left\\(\\text{mod}\\left(\\text{floor}\\left(\\frac{1}{p}\\right)\\right)\\right)$.","description_state":"present","introduced_year":null,"introduced_by":{"title":"Reducing Transformer Depth on Demand with Structured Dropout","paper":"/paper/reducing-transformer-depth-on-demand-with-1","first_author":"Angela Fan","n_authors":3,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/reducing-transformer-depth-on-demand-with-1"},"source":{"url":"https://arxiv.org/abs/1909.11556v1","title":"Reducing Transformer Depth on Demand with Structured Dropout","url_on_a_paper_host":true},"code_snippet_url":"https://github.com/pytorch/fairseq/blob/9ebcd6554daab7bac4948de49aeb85bfff81d876/fairseq/checkpoint_utils.py#L365-L446","code_snippet_url_on_a_code_host":true,"categories":[{"area":"General","area_id":"general","collection":"Regularization","url":"/methods/category/regularization","pwc_aliases":[]}],"n_papers_tagged":4,"archive_num_papers":4,"papers_newest_first":[{"paper":"/paper/layer-skip-enabling-early-exit-inference-and","title":"LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding","date":"2024-04-25","arxiv_id":"2404.16710","n_code_links":1,"syntology":{"ran":4,"of":5,"unverified":1,"pointer_only":5}},{"paper":"/paper/perceiver-vl-efficient-vision-and-language","title":"Perceiver-VL: Efficient Vision-and-Language Modeling with Iterative Latent Attention","date":"2022-11-21","arxiv_id":"2211.11701","n_code_links":1,"syntology":null},{"paper":null,"title":"Training Flexible Depth Model by Multi-Task Learning for Neural Machine Translation","date":"2020-10-16","arxiv_id":"2010.08265","n_code_links":0,"syntology":null},{"paper":"/paper/reducing-transformer-depth-on-demand-with-1","title":"Reducing Transformer Depth on Demand with Structured Dropout","date":"2019-09-25","arxiv_id":"1909.11556","n_code_links":5,"syntology":null}],"papers_shown":4,"tasks":[{"task":"/task/language-modeling","name":"Language Modeling","papers":2},{"task":"/task/language-modelling","name":"Language Modelling","papers":2},{"task":"/task/machine-translation","name":"Machine Translation","papers":2},{"task":"/task/translation","name":"Translation","papers":2},{"task":"/task/cross-modal-retrieval","name":"Cross-Modal Retrieval","papers":1},{"task":"/task/gsm8k","name":"GSM8K","papers":1},{"task":"/task/hellaswag","name":"HellaSwag","papers":1},{"task":"/task/humaneval","name":"HumanEval","papers":1},{"task":"/task/mmlu","name":"MMLU","papers":1},{"task":"/task/math","name":"Math","papers":1},{"task":"/task/multi-task-learning","name":"Multi-Task Learning","papers":1},{"task":"/task/natural-questions","name":"Natural Questions","papers":1},{"task":"/task/open-domain-question-answering","name":"Open-Domain Question Answering","papers":1},{"task":"/task/question-answering","name":"Question Answering","papers":1},{"task":"/task/race-h","name":"RACE-h","papers":1},{"task":"/task/race-m","name":"RACE-m","papers":1},{"task":"/task/retrieval","name":"Retrieval","papers":1},{"task":"/task/triviaqa","name":"TriviaQA","papers":1},{"task":"/task/winogrande","name":"Winogrande","papers":1}],"tasks_shown":19,"n_tasks":19,"usage_by_year":[{"year":"2019","papers":1},{"year":"2020","papers":1},{"year":"2022","papers":1},{"year":"2024","papers":1}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/layerdrop"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}