Mixture of Attentions For Speculative Decoding

• URL: http://arxiv.org/abs/2410.03804v1
• Date: Fri, 4 Oct 2024 10:25:52 GMT
• Title: Mixture of Attentions For Speculative Decoding
• Authors: Matthieu Zimmer, Milan Gritta, Gerasimos Lampouras, Haitham Bou Ammar, Jun Wang,
• Abstract summary: Speculative decoding (SD) leverages smaller models to efficiently propose future tokens, which are then verified by the Large Language Models in parallel. We identify several limitations of SD models including the lack of on-policyness during training and partial observability. We propose a more grounded architecture for small models by introducing a Mixture of Attentions for SD.
• Score: 17.344416130742232
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The growth in the number of parameters of Large Language Models (LLMs) has led to a significant surge in computational requirements, making them challenging and costly to deploy. Speculative decoding (SD) leverages smaller models to efficiently propose future tokens, which are then verified by the LLM in parallel. Small models that utilise activations from the LLM currently achieve the fastest decoding speeds. However, we identify several limitations of SD models including the lack of on-policyness during training and partial observability. To address these shortcomings, we propose a more grounded architecture for small models by introducing a Mixture of Attentions for SD. Our novel architecture can be applied in two scenarios: a conventional single device deployment and a novel client-server deployment where the small model is hosted on a consumer device and the LLM on a server. In a single-device scenario, we demonstrate state-of-the-art speedups improving EAGLE-2 by 9.5% and its acceptance length by 25%. In a client-server setting, our experiments demonstrate: 1) state-of-the-art latencies with minimal calls to the server for different network conditions, and 2) in the event of a complete disconnection, our approach can maintain higher accuracy compared to other SD methods and demonstrates advantages over API calls to LLMs, which would otherwise be unable to continue the generation process.

Related papers

• Read-ME: Refactorizing LLMs as Router-Decoupled Mixture of Experts with System Co-Design [59.00758127310582]
  We propose a novel framework Read-ME that transforms pre-trained dense LLMs into smaller MoE models. Our approach employs activation sparsity to extract experts. Read-ME outperforms other popular open-source dense models of similar scales.
  arXiv  Detail & Related papers  (2024-10-24T19:48:51Z)
• Dual-Model Distillation for Efficient Action Classification with Hybrid Edge-Cloud Solution [1.8029479474051309]
  We design a hybrid edge-cloud solution that leverages the efficiency of smaller models for local processing while deferring to larger, more accurate cloud-based models when necessary. Specifically, we propose a novel unsupervised data generation method, Dual-Model Distillation (DMD), to train a lightweight switcher model that can predict when the edge model's output is uncertain. Experimental results on the action classification task show that our framework not only requires less computational overhead, but also improves accuracy compared to using a large model alone.
  arXiv  Detail & Related papers  (2024-10-16T02:06:27Z)
• SWIFT: On-the-Fly Self-Speculative Decoding for LLM Inference Acceleration [10.970637831760136]
  Speculative decoding (SD) has emerged as a widely used paradigm to accelerate the inference of large language models (LLMs) We introduce SWIFT, an on-the-fly self-speculative decoding algorithm that adaptively selects intermediate layers of LLMs to skip during inference. We show that SWIFT can achieve over a 1.3x-1.6x speedup while preserving the original distribution of the generated text.
  arXiv  Detail & Related papers  (2024-10-09T14:15:30Z)
• Reference Trustable Decoding: A Training-Free Augmentation Paradigm for Large Language Models [79.41139393080736]
  Large language models (LLMs) have rapidly advanced and demonstrated impressive capabilities. We propose Reference Trustable Decoding (RTD), a paradigm that allows models to quickly adapt to new tasks without fine-tuning.
  arXiv  Detail & Related papers  (2024-09-30T10:48:20Z)
• ELMS: Elasticized Large Language Models On Mobile Devices [5.689405542579458]
  On-device Large Language Models (LLMs) are revolutionizing mobile AI, enabling applications such as UI automation while addressing privacy concerns. We introduce ELMS, an on-device LLM service designed to provide elasticity in both the model and prompt dimensions. A one-time reorder neuroning technique, which utilizes the inherent permutation consistency within transformer models to create high-quality, elastic sub-models. A dual-head compact language model, which efficiently refines prompts and coordinates the elastic adaptation between the model prompt.
  arXiv  Detail & Related papers  (2024-09-08T06:32:08Z)
• Delta-CoMe: Training-Free Delta-Compression with Mixed-Precision for Large Language Models [79.46938238953916]
  Fine-tuning large language models (LLMs) to diverse applications is crucial to meet complex demands. Recent studies suggest decomposing a fine-tuned LLM into a base model and corresponding delta weights, which are then compressed using low-rank or low-bit approaches to reduce costs. In this work, we observe that existing low-rank and low-bit compression methods can significantly harm the model performance for task-specific fine-tuned LLMs.
  arXiv  Detail & Related papers  (2024-06-13T07:57:27Z)
• Distributed Inference and Fine-tuning of Large Language Models Over The Internet [91.00270820533272]
  Large language models (LLMs) are useful in many NLP tasks and become more capable with size. These models require high-end hardware, making them inaccessible to most researchers. We develop fault-tolerant inference algorithms and load-balancing protocols that automatically assign devices to maximize the total system throughput.
  arXiv  Detail & Related papers  (2023-12-13T18:52:49Z)
• Federated Full-Parameter Tuning of Billion-Sized Language Models with Communication Cost under 18 Kilobytes [53.4856038354195]
  Pre-trained large language models (LLMs) need fine-tuning to improve their responsiveness to natural language instructions. FedKSeed employs zeroth-order optimization with a finite set of random seeds. It significantly reduces transmission requirements between the server and clients to just a few random seeds.
  arXiv  Detail & Related papers  (2023-12-11T13:03:21Z)
• Enhancing Neural Architecture Search with Multiple Hardware Constraints for Deep Learning Model Deployment on Tiny IoT Devices [17.919425885740793]
  We propose a novel approach to incorporate multiple constraints into so-called Differentiable NAS optimization methods. We show that, with a single search, it is possible to reduce memory and latency by 87.4% and 54.2%, respectively.
  arXiv  Detail & Related papers  (2023-10-11T06:09:14Z)
• PFSL: Personalized & Fair Split Learning with Data & Label Privacy for thin clients [0.5144809478361603]
  PFSL is a new framework of distributed split learning where a large number of thin clients perform transfer learning in parallel. We implement a lightweight step of personalization of client models to provide high performance for their respective data distributions. Our accuracy far exceeds that of current algorithms SL and is very close to that of centralized learning on several real-life benchmarks.
  arXiv  Detail & Related papers  (2023-03-19T10:38:29Z)
• Speculative Decoding with Big Little Decoder [108.95187338417541]
  Big Little Decoder (BiLD) is a framework that can improve inference efficiency and latency for a wide range of text generation applications. On an NVIDIA T4 GPU, our framework achieves a speedup of up to 2.12x speedup with minimal generation quality degradation. Our framework is fully plug-and-play and can be applied without any modifications in the training process or model architecture.
  arXiv  Detail & Related papers  (2023-02-15T18:55:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.