Break the sequential dependency of llm inference using lookahead decoding

arXiv preprint arXiv:2401.09670

DistServe improves the performance of large language models (LLMs) serving by disaggregating the prefill and decoding computation. Existing LLM serving systems colocate the two phases and batch the computation of prefill and decoding across all users and requests. We find that this strategy not only leads to strong prefill-decoding interferences but also couples the resource allocation and parallelism plans for both phases. LLM applications often emphasize individual latency for each phase: time to first token (TTFT) for the prefill phase and time per output token (TPOT) of each request for the decoding phase. In the presence of stringent latency requirements, existing systems have to prioritize one latency over the other, or over-provision compute resources to meet both. DistServe assigns prefill and decoding computation to different GPUs, hence eliminating prefill-decoding interferences. Given the application's TTFT and TPOT requirements, DistServe co-optimizes the resource allocation and parallelism strategy tailored for each phase. DistServe also places the two phases according to the serving cluster's bandwidth to minimize the communication caused by disaggregation. As a result, DistServe significantly improves LLM serving performance in terms of the maximum rate that can be served within both TTFT and TPOT constraints on each GPU. Our evaluations show that on various popular LLMs, applications, and latency requirements, DistServe can serve 4.48x more requests or 10.2x tighter SLO, compared to state-of-the-art systems, while staying within latency constraints for > 90% of requests.

arXiv preprint arXiv:2403.13839

PyTorch \texttt{2.x} introduces a compiler designed to accelerate deep learning programs. However, for machine learning researchers, adapting to the PyTorch compiler to full potential can be challenging. The compiler operates at the Python bytecode level, making it appear as an opaque box. To address this, we introduce \texttt{depyf}, a tool designed to demystify the inner workings of the PyTorch compiler. \texttt{depyf} decompiles bytecode generated by PyTorch back into equivalent source code, and establishes connections between in-memory code objects and their on-disk source code counterparts. This feature enables users to step through the source code line by line using debuggers, thus enhancing their understanding of the underlying processes. Notably, \texttt{depyf} is non-intrusive and user-friendly, primarily relying on two convenient context managers for its core functionality. The project is \href{https://github.com/thuml/depyf}{ openly available} and is recognized as a \href{https://pytorch.org/ecosystem/}{PyTorch ecosystem project}.

arXiv preprint arXiv:2404.14527

Large language models (LLMs) are increasingly integrated into many online services. However, a major challenge in deploying LLMs is their high cost, due primarily to the use of expensive GPU instances. To address this problem, we find that the significant heterogeneity of GPU types presents an opportunity to increase GPU cost efficiency and reduce deployment costs. The broad and growing market of GPUs creates a diverse option space with varying costs and hardware specifications. Within this space, we show that there is not a linear relationship between GPU cost and performance, and identify three key LLM service characteristics that significantly affect which GPU type is the most cost effective: model request size, request rate, and latency service-level objective (SLO). We then present M\'elange, a framework for navigating the diversity of GPUs and LLM service specifications to derive the most cost-efficient set of GPUs for a given LLM service. We frame the task of GPU selection as a cost-aware bin-packing problem, where GPUs are bins with a capacity and cost, and items are request slices defined by a request size and rate. Upon solution, M\'elange derives the minimal-cost GPU allocation that adheres to a configurable latency SLO. Our evaluations across both real-world and synthetic datasets demonstrate that M\'elange can reduce deployment costs by up to 77% as compared to utilizing only a single GPU type, highlighting the importance of making heterogeneity-aware GPU provisioning decisions for LLM serving. Our source code is publicly available at https://github.com/tyler-griggs/melange-release.

NSDI 2024

Cloud providers offer spot instances alongside on-demand instances to optimize resource utilization. While economically appealing, spot instances’ preemptible nature causes them ill-suited for deadline-sensitive jobs. To allow jobs to meet deadlines while leveraging spot instances, we propose a simple idea: use on-demand instances judiciously as a backup resource. However, due to the unpredictable spot instance availability, determining when to switch between spot and on-demand to minimize cost requires careful policy design. In this paper, we first provide an in-depth characterization of spot instances (eg, availability, pricing, duration), and develop a basic theoretical model to examine the worst and average-case behaviors of baseline policies (eg, greedy). The model serves as a foundation to motivate our design of a simple and effective policy, Uniform Progress, which is parameter-free and requires no …

arXiv preprint arXiv:2403.07974

Large Language Models (LLMs) applied to code-related applications have emerged as a prominent field, attracting significant interest from both academia and industry. However, as new and improved LLMs are developed, existing evaluation benchmarks (e.g., HumanEval, MBPP) are no longer sufficient for assessing their capabilities. In this work, we propose LiveCodeBench, a comprehensive and contamination-free evaluation of LLMs for code, which continuously collects new problems over time from contests across three competition platforms, namely LeetCode, AtCoder, and CodeForces. Notably, our benchmark also focuses on a broader range of code related capabilities, such as self-repair, code execution, and test output prediction, beyond just code generation. Currently, LiveCodeBench hosts four hundred high-quality coding problems that were published between May 2023 and February 2024. We have evaluated 9 base LLMs and 20 instruction-tuned LLMs on LiveCodeBench. We present empirical findings on contamination, holistic performance comparisons, potential overfitting in existing benchmarks as well as individual model comparisons. We will release all prompts and model completions for further community analysis, along with a general toolkit for adding new scenarios and model

Machine learning (ML) is increasingly used behind closed systems and APIs to make important decisions. For example, social media uses ML-based recommendation algorithms to decide what to show users, and millions of people pay to use ChatGPT for information every day. Because ML is deployed behind these closed systems, there are increasing calls for transparency, such as releasing model weights. However, these service providers have legitimate reasons not to release this information, including for privacy and trade secrets. To bridge this gap, recent work has proposed using zero-knowledge proofs (specifically a form called ZK-SNARKs) for certifying computation with private models but has only been applied to unrealistically small models. In this work, we present the first framework, ZKML, to produce ZK-SNARKs for realistic ML models, including state-of-the-art vision models, a distilled GPT-2, and the …

Advances in Neural Information Processing Systems

Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80\% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github. com/lm-sys/FastChat/tree/main/fastchat/llm_judge.

arXiv preprint arXiv:2404.02015

Large language models (LLMs) have demon- strated remarkable performance, and organiza- tions are racing to serve LLMs of varying sizes as endpoints for use-cases like chat, programming and search. However, efficiently serving multiple LLMs poses significant challenges for existing approaches due to varying popularity of LLMs. In the paper, we present MuxServe, a flexible spatial-temporal multiplexing system for efficient multiple LLM serving. The key insight behind is to colocate LLMs considering their popularity to multiplex memory resources, and leverage the characteristics of prefill and decoding phases to separate and flexibly colocate them to multiplex computation resources. MuxServe formally for- mulates the multiplexing problem, and proposes a novel placement algorithm and adaptive batch scheduling strategy to identify optimal coloca- tions and maximize utilization. MuxServe de- signs a unified resource manager to enable flexi- ble and efficient multiplexing. Evaluation results show that MuxServe can achieves up to higher throughput or processes more requests within SLO attainment.

Bulk data replication across multiple cloud regions and providers is essential for large organizations to support data analytics, disaster recovery, and geo-distributed model serving. However, data multicast in the cloud can be expensive due to network egress costs and slow due to cloud network constraints. In this paper, we study the design of high-throughput, cost-optimized overlay multicast for bulk cloud data replication that exploits trends in modern provider pricing models along with techniques like ephemeral waypoints to minimize cloud networking costs.

arXiv preprint arXiv:2403.04132

Large Language Models (LLMs) have unlocked new capabilities and applications; however, evaluating the alignment with human preferences still poses significant challenges. To address this issue, we introduce Chatbot Arena, an open platform for evaluating LLMs based on human preferences. Our methodology employs a pairwise comparison approach and leverages input from a diverse user base through crowdsourcing. The platform has been operational for several months, amassing over 240K votes. This paper describes the platform, analyzes the data we have collected so far, and explains the tried-and-true statistical methods we are using for efficient and accurate evaluation and ranking of models. We confirm that the crowdsourced questions are sufficiently diverse and discriminating and that the crowdsourced human votes are in good agreement with those of expert raters. These analyses collectively establish a robust foundation for the credibility of Chatbot Arena. Because of its unique value and openness, Chatbot Arena has emerged as one of the most referenced LLM leaderboards, widely cited by leading LLM developers and companies. Our demo is publicly available at \url{https://chat.lmsys.org}.

arXiv preprint arXiv:2404.06921

Large Language Models (LLMs) are evolving beyond their classical role of providing information within dialogue systems to actively engaging with tools and performing actions on real-world applications and services. Today, humans verify the correctness and appropriateness of the LLM-generated outputs (e.g., code, functions, or actions) before putting them into real-world execution. This poses significant challenges as code comprehension is well known to be notoriously difficult. In this paper, we study how humans can efficiently collaborate with, delegate to, and supervise autonomous LLMs in the future. We argue that in many cases, "post-facto validation" - verifying the correctness of a proposed action after seeing the output - is much easier than the aforementioned "pre-facto validation" setting. The core concept behind enabling a post-facto validation system is the integration of an intuitive undo feature, and establishing a damage confinement for the LLM-generated actions as effective strategies to mitigate the associated risks. Using this, a human can now either revert the effect of an LLM-generated output or be confident that the potential risk is bounded. We believe this is critical to unlock the potential for LLM agents to interact with applications and services with limited (post-facto) human involvement. We describe the design and implementation of our open-source runtime for executing LLM actions, Gorilla Execution Engine (GoEX), and present open research questions towards realizing the goal of LLMs and applications interacting with each other with minimal human supervision. We release GoEX at https://github.com/ShishirPatil/gorilla/.

arXiv preprint arXiv:2403.02419

Many recent state-of-the-art results in language tasks were achieved using compound systems that perform multiple Large Language Model (LLM) calls and aggregate their responses. However, there is little understanding of how the number of LLM calls -- e.g., when asking the LLM to answer each question multiple times and taking a consensus -- affects such a compound system's performance. In this paper, we initiate the study of scaling laws of compound inference systems. We analyze, theoretically and empirically, how the number of LLM calls affects the performance of one-layer Voting Inference Systems -- one of the simplest compound systems, which aggregates LLM responses via majority voting. We find empirically that across multiple language tasks, surprisingly, Voting Inference Systems' performance first increases but then decreases as a function of the number of LLM calls. Our theoretical results suggest that this non-monotonicity is due to the diversity of query difficulties within a task: more LLM calls lead to higher performance on "easy" queries, but lower performance on "hard" queries, and non-monotone behavior emerges when a task contains both types of queries. This insight then allows us to compute, from a small number of samples, the number of LLM calls that maximizes system performance, and define a scaling law of Voting Inference Systems. Experiments show that our scaling law can predict the performance of Voting Inference Systems and find the optimal number of LLM calls to make.

arXiv preprint arXiv:2404.04500

There is an increasing conflict between business incentives to hide models and data as trade secrets, and the societal need for algorithmic transparency. For example, a rightsholder wishing to know whether their copyrighted works have been used during training must convince the model provider to allow a third party to audit the model and data. Finding a mutually agreeable third party is difficult, and the associated costs often make this approach impractical. In this work, we show that it is possible to simultaneously allow model providers to keep their model weights (but not architecture) and data secret while allowing other parties to trustlessly audit model and data properties. We do this by designing a protocol called ZkAudit in which model providers publish cryptographic commitments of datasets and model weights, alongside a zero-knowledge proof (ZKP) certifying that published commitments are derived from training the model. Model providers can then respond to audit requests by privately computing any function F of the dataset (or model) and releasing the output of F alongside another ZKP certifying the correct execution of F. To enable ZkAudit, we develop new methods of computing ZKPs for SGD on modern neural nets for simple recommender systems and image classification models capable of high accuracies on ImageNet. Empirically, we show it is possible to provide trustless audits of DNNs, including copyright, censorship, and counterfactual audits with little to no loss in accuracy.

arXiv preprint arXiv:2404.02852

Mixture-of-Expert (MoE) based large language models (LLMs), such as the recent Mixtral and DeepSeek-MoE, have shown great promise in scaling model size without suffering from the quadratic growth of training cost of dense transformers. Like dense models, training MoEs requires answering the same question: given a training budget, what is the optimal allocation on the model size and number of tokens? We study the scaling law of MoE-based LLMs regarding the relations between the model performance, model size, dataset size, and the expert degree. Echoing previous research studying MoE in different contexts, we observe the diminishing return of increasing the number of experts, but this seems to suggest we should scale the number of experts until saturation, as the training cost would remain constant, which is problematic during inference time. We propose to amend the scaling law of MoE by introducing inference efficiency as another metric besides the validation loss. We find that MoEs with a few (4/8) experts are the most serving efficient solution under the same performance, but costs 2.5-3.5x more in training. On the other hand, training a (16/32) expert MoE much smaller (70-85%) than the loss-optimal solution, but with a larger training dataset is a promising setup under a training budget.

arXiv preprint arXiv:2402.01869

Large language models are increasingly integrated with external tools and APIs like ChatGPT plugins to extend their capability beyond language-centric tasks. However, today's LLM inference systems are designed for standalone LLMs. They treat API calls as new requests, causing unnecessary recomputation of already computed contexts, which accounts for 37-40% of total model forwarding time. This paper presents APIServe, the first LLM inference framework targeting API-augmented LLMs. APISERVE minimizes the GPU resource waste caused by API calls and dedicates saved memory for serving more requests. APISERVE improves the overall serving throughput by 1.6x and completes 2x more requests per second compared to the state-of-the-art LLM inference systems.

arXiv preprint arXiv:2403.04132

Large Language Models (LLMs) have unlocked new capabilities and applications; however, evaluating the alignment with human preferences still poses significant challenges. To address this issue, we introduce Chatbot Arena, an open platform for evaluating LLMs based on human preferences. Our methodology employs a pairwise comparison approach and leverages input from a diverse user base through crowdsourcing. The platform has been operational for several months, amassing over 240K votes. This paper describes the platform, analyzes the data we have collected so far, and explains the tried-and-true statistical methods we are using for efficient and accurate evaluation and ranking of models. We confirm that the crowdsourced questions are sufficiently diverse and discriminating and that the crowdsourced human votes are in good agreement with those of expert raters. These analyses collectively establish a robust foundation for the credibility of Chatbot Arena. Because of its unique value and openness, Chatbot Arena has emerged as one of the most referenced LLM leaderboards, widely cited by leading LLM developers and companies. Our demo is publicly available at \url{https://chat.lmsys.org}.

Advances in Neural Information Processing Systems

Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80\% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github. com/lm-sys/FastChat/tree/main/fastchat/llm_judge.

arXiv preprint arXiv:2404.02015

Large language models (LLMs) have demon- strated remarkable performance, and organiza- tions are racing to serve LLMs of varying sizes as endpoints for use-cases like chat, programming and search. However, efficiently serving multiple LLMs poses significant challenges for existing approaches due to varying popularity of LLMs. In the paper, we present MuxServe, a flexible spatial-temporal multiplexing system for efficient multiple LLM serving. The key insight behind is to colocate LLMs considering their popularity to multiplex memory resources, and leverage the characteristics of prefill and decoding phases to separate and flexibly colocate them to multiplex computation resources. MuxServe formally for- mulates the multiplexing problem, and proposes a novel placement algorithm and adaptive batch scheduling strategy to identify optimal coloca- tions and maximize utilization. MuxServe de- signs a unified resource manager to enable flexi- ble and efficient multiplexing. Evaluation results show that MuxServe can achieves up to higher throughput or processes more requests within SLO attainment.

arXiv preprint arXiv:2402.02057

Autoregressive decoding of large language models (LLMs) is memory bandwidth bounded, resulting in high latency and significant wastes of the parallel processing power of modern accelerators. Existing methods for accelerating LLM decoding often require a draft model (e.g., speculative decoding), which is nontrivial to obtain and unable to generalize. In this paper, we introduce Lookahead decoding, an exact, parallel decoding algorithm that accelerates LLM decoding without needing auxiliary models or data stores. It allows trading per-step log(FLOPs) to reduce the number of total decoding steps, is more parallelizable on single or multiple modern accelerators, and is compatible with concurrent memory-efficient attention (e.g., FlashAttention). Our implementation of Lookahead decoding can speed up autoregressive decoding by up to 1.8x on MT-bench and 4x with strong scaling on multiple GPUs in code completion tasks. Our code is avialable at https://github.com/hao-ai-lab/LookaheadDecoding

arXiv preprint arXiv:2402.02057

Autoregressive decoding of large language models (LLMs) is memory bandwidth bounded, resulting in high latency and significant wastes of the parallel processing power of modern accelerators. Existing methods for accelerating LLM decoding often require a draft model (e.g., speculative decoding), which is nontrivial to obtain and unable to generalize. In this paper, we introduce Lookahead decoding, an exact, parallel decoding algorithm that accelerates LLM decoding without needing auxiliary models or data stores. It allows trading per-step log(FLOPs) to reduce the number of total decoding steps, is more parallelizable on single or multiple modern accelerators, and is compatible with concurrent memory-efficient attention (e.g., FlashAttention). Our implementation of Lookahead decoding can speed up autoregressive decoding by up to 1.8x on MT-bench and 4x with strong scaling on multiple GPUs in code completion tasks. Our code is avialable at https://github.com/hao-ai-lab/LookaheadDecoding