eprintid: 10195822 rev_number: 7 eprint_status: archive userid: 699 dir: disk0/10/19/58/22 datestamp: 2024-08-16 11:56:11 lastmod: 2024-08-16 11:56:11 status_changed: 2024-08-16 11:56:11 type: proceedings_section metadata_visibility: show sword_depositor: 699 creators_name: Shi, Z creators_name: Lipani, A title: DePT: Decomposed Prompt Tuning for Parameter-Efficient Fine-tuning ispublished: pub divisions: UCL divisions: B04 divisions: F44 keywords: Natural Language Processing, Large Language Models, Parameter-efficient Fine-tuning note: This version is the version of record. For information on re-use, please refer to the publisher’s terms and conditions. abstract: Prompt tuning (PT), where a small amount of trainable soft (continuous) prompt vectors is affixed to the model input, has shown promising results across various tasks and model architecture for parameter-efficient fine-tuning (PEFT). PT stands out from other PEFT approaches because it maintains competitive performance with fewer trainable parameters and does not drastically scale up its parameters as the model size expands. However, PT introduces extra soft prompt tokens, leading to longer input sequences, which significantly impacts training/inference time and memory usage due to the Transformer's quadratic complexity. Particularly concerning for Large Language Models (LLMs) that face heavy daily querying. To address this issue, we propose Decomposed Prompt Tuning (DEPT), which decomposes the soft prompt into a shorter soft prompt and a pair of low-rank matrices that are then optimised with two different learning rates. This allows DEPT to achieve better performance while saving substantial memory and time costs compared to vanilla PT and its variants, without changing trainable parameter sizes. Through extensive experiments on 23 natural language processing (NLP) and vision-language (VL) tasks, we demonstrate that DEPT outperforms state-of-the-art PEFT approaches, including the full fine-tuning baseline, in some scenarios. Additionally, we empirically show that DEPT grows more efficient as the model size increases. Our further study reveals that DEPT integrates seamlessly with parameter-efficient transfer learning in the few-shot learning setting and highlights its adaptability to various model architectures and sizes. date: 2024-05-11 date_type: published publisher: International Conference on Learning Representations (ICLR) official_url: https://openreview.net/forum?id=KjegfPGRde oa_status: green full_text_type: pub language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 2305017 lyricists_name: Lipani, Aldo lyricists_id: ALIPA33 actors_name: Flynn, Bernadette actors_id: BFFLY94 actors_role: owner full_text_status: public pres_type: paper series: ICLR publication: 12th International Conference on Learning Representations, ICLR 2024 volume: 2024 place_of_pub: Vienna, Austria event_title: 12th International Conference on Learning Representations, ICLR 2024 book_title: 12th International Conference on Learning Representations, ICLR 2024 citation: Shi, Z; Lipani, A; (2024) DePT: Decomposed Prompt Tuning for Parameter-Efficient Fine-tuning. In: 12th International Conference on Learning Representations, ICLR 2024. International Conference on Learning Representations (ICLR): Vienna, Austria. Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10195822/1/725_DePT_Decomposed_Prompt_Tun.pdf