paper-with-me

홈 › Papers

Fine-tuning Language Models over Slow Networks using Activation Compression with Guarantees

2022-06-02 · Jue Wang, Binhang Yuan, Luka Rimanic, Yongjun He, Tri Dao, Beidi Chen, Christopher Re, Ce Zhang

Communication compression is a crucial technique for modern distributed learning systems to alleviate their communication bottlenecks over slower networks. Despite recent intensive studies of gradient compression for data parallel-style training, compressing the activations for models trained with pipeline parallelism is still an open problem. In this paper, we propose AC-SGD, a novel activation compression algorithm for communication-efficient pipeline parallelism training over slow networks. Different from previous efforts in activation compression, instead of compressing activation values directly, AC-SGD compresses the changes of the activations. This allows us to show, to the best of our knowledge for the first time, that one can still achieve $O(1/\sqrt{T})$ convergence rate for non-convex objectives under activation compression, without making assumptions on gradient unbiasedness that do not hold for deep learning models with non-linear activation functions.We then show that AC-SGD can be optimized and implemented efficiently, without additional end-to-end runtime overhead.We evaluated AC-SGD to fine-tune language models with up to 1.5 billion parameters, compressing activations to 2-4 bits.AC-SGD provides up to 4.3X end-to-end speed-up in slower networks, without sacrificing model quality. Moreover, we also show that AC-SGD can be combined with state-of-the-art gradient compression algorithms to enable "end-to-end communication compression: All communications between machines, including model gradients, forward activations, and backward gradients are compressed into lower precision.This provides up to 4.9X end-to-end speed-up, without sacrificing model quality.

📄 PDF Abstract BibTeX arXiv:2206.01299

Code (1)

DS3Lab/AC-SGD 공식 구현 pytorch

Similar Papers 제목 키워드 기반

Src Kinase Slows Collective Rotation of Confined Epithelial Cell Monolayers

2024-07-09 · Nastassia Pricoupenko, Flavia Marsigliesi, Philippe Marcq, Carles Blanch-Mercader 외

Collective cell migration is key during development, wound healing and metastasis and relies on coordinated cell behaviors at the group level. Src kinase is a key signalling protein for physiological functions of epithel…

Friction

Thinking Forward: Memory-Efficient Federated Finetuning of Language Models

2024-05-24 · Kunjal Panchal, Nisarg Parikh, Sunav Choudhary, Lijun Zhang 외

Finetuning large language models (LLMs) in federated learning (FL) settings has become increasingly important as it allows resource-constrained devices to finetune a model using private data. However, finetuning LLMs usi…

Federated Learning

LoRA-FA: Memory-efficient Low-rank Adaptation for Large Language Models Fine-tuning

2023-08-07 · Longteng Zhang, Lin Zhang, Shaohuai Shi, Xiaowen Chu 외

The low-rank adaptation (LoRA) method can largely reduce the amount of trainable parameters for fine-tuning large language models (LLMs), however, it still requires expensive activation memory to update low-rank weights.…

Analyzing and Reducing the Performance Gap in Cross-Lingual Transfer with Fine-tuning Slow and Fast

2023-05-19 · Yiduo Guo, Yaobo Liang, Dongyan Zhao, Bing Liu 외

Existing research has shown that a multilingual pre-trained language model fine-tuned with one (source) language also performs well on downstream tasks for non-source languages, even though no fine-tuning is done on thes…

Cross-Lingual TransferLanguage ModelingLanguage Modelling

Mitigating Outlier Activations in Low-Precision Fine-Tuning of Language Models

2023-12-14 · Alireza Ghaffari, Justin Yu, Mahsa Ghazvini Nejad, Masoud Asgharian 외

Low-precision fine-tuning of language models has gained prominence as a cost-effective and energy-efficient approach to deploying large-scale models in various applications. However, this approach is susceptible to the e…