Guardado en:
Detalles Bibliográficos
Autores principales: Pentyala, Shiva Kumar, Wang, Zhichao, Bi, Bin, Ramnath, Kiran, Mao, Xiang-Bo, Radhakrishnan, Regunathan, Asur, Sitaram, Na, Cheng
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2406.17923
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910502119014400
author Pentyala, Shiva Kumar
Wang, Zhichao
Bi, Bin
Ramnath, Kiran
Mao, Xiang-Bo
Radhakrishnan, Regunathan
Asur, Sitaram
Na
Cheng
author_facet Pentyala, Shiva Kumar
Wang, Zhichao
Bi, Bin
Ramnath, Kiran
Mao, Xiang-Bo
Radhakrishnan, Regunathan
Asur, Sitaram
Na
Cheng
contents Large language models (LLMs) have shown remarkable abilities in diverse natural language processing (NLP) tasks. The LLMs generally undergo supervised fine-tuning (SFT) followed by preference alignment to be usable in downstream applications. However, this sequential training pipeline leads to alignment tax that degrades the LLM performance. This paper introduces PAFT, a new PArallel training paradigm for effective LLM Fine-Tuning, which independently performs SFT and preference alignment (e.g., DPO and ORPO, etc.) with the same pre-trained model on respective datasets. The model produced by SFT and the model from preference alignment are then merged into a final model by parameter fusing for use in downstream applications. This work reveals important findings that preference alignment like DPO naturally results in a sparse model while SFT leads to a natural dense model which needs to be sparsified for effective model merging. This paper introduces an effective interference resolution which reduces the redundancy by sparsifying the delta parameters. The LLM resulted from the new training paradigm achieved Rank #1 on the HuggingFace Open LLM Leaderboard. Comprehensive evaluation shows the effectiveness of the parallel training paradigm.
format Preprint
id arxiv_https___arxiv_org_abs_2406_17923
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle PAFT: A Parallel Training Paradigm for Effective LLM Fine-Tuning
Pentyala, Shiva Kumar
Wang, Zhichao
Bi, Bin
Ramnath, Kiran
Mao, Xiang-Bo
Radhakrishnan, Regunathan
Asur, Sitaram
Na
Cheng
Computation and Language
Large language models (LLMs) have shown remarkable abilities in diverse natural language processing (NLP) tasks. The LLMs generally undergo supervised fine-tuning (SFT) followed by preference alignment to be usable in downstream applications. However, this sequential training pipeline leads to alignment tax that degrades the LLM performance. This paper introduces PAFT, a new PArallel training paradigm for effective LLM Fine-Tuning, which independently performs SFT and preference alignment (e.g., DPO and ORPO, etc.) with the same pre-trained model on respective datasets. The model produced by SFT and the model from preference alignment are then merged into a final model by parameter fusing for use in downstream applications. This work reveals important findings that preference alignment like DPO naturally results in a sparse model while SFT leads to a natural dense model which needs to be sparsified for effective model merging. This paper introduces an effective interference resolution which reduces the redundancy by sparsifying the delta parameters. The LLM resulted from the new training paradigm achieved Rank #1 on the HuggingFace Open LLM Leaderboard. Comprehensive evaluation shows the effectiveness of the parallel training paradigm.
title PAFT: A Parallel Training Paradigm for Effective LLM Fine-Tuning
topic Computation and Language
url https://arxiv.org/abs/2406.17923