NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference

• URL: http://arxiv.org/abs/2411.01142v1
• Date: Sat, 02 Nov 2024 05:15:44 GMT
• Title: NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference
• Abstract summary: NEO is an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU. We present NEO, an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU.
• Score: 22.24693513075852
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Online LLM inference powers many exciting applications such as intelligent chatbots and autonomous agents. Modern LLM inference engines widely rely on request batching to improve inference throughput, aiming to make it cost-efficient when running on expensive GPU accelerators. However, the limited GPU memory has largely limited the batch size achieved in practice, leaving significant GPU compute resources wasted. We present NEO, an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU, effectively increasing the GPU batch size and thus inference throughput. To this end, NEO proposes asymmetric GPU-CPU pipelining and load-aware scheduling to balance GPU and CPU loads and fully utilize their compute and memory resources. We evaluate NEO on a wide range of workloads (i.e., code generation, text summarization), GPUs (i.e., T4, A10G, H100), and LLM models (i.e., 7B, 8B, 70B). NEO achieves up to 7.5$\times$, 26%, and 14% higher throughput compared to GPU-only approach on T4, A10G, and H100 GPUs, respectively, while maintaining the same latency; with more powerful CPUs, NEO achieves up to 79.3% throughput gain on A10G GPU.

Related papers

• Mixture-of-Parallelisms: Towards Memory-Efficient Training Stack for Mixture-of-Experts Models [67.17268530365189]
  This paper showcases a memory-efficient training stack for Mixture-of-Experts (MoE) models.<n>It combines and specializes various existing and novel parallelism techniques at different layers and stages of the MoE model training pipeline.
  arXiv  Detail & Related papers  (2026-07-02T08:06:57Z)
• GPU Acceleration of Learning With Errors KEMs Using OpenACC for Post-Quantum Cryptography [0.0]
  Key Encapsulation Mechanism (KEM) based on plain Learning with Errors (LWE) problem.<n>Implementation achieves significant acceleration across all tested GPU platforms.<n>Results highlight the effectiveness of GPU acceleration for computationally demanding LWE-based cryptographic workloads.
  arXiv  Detail & Related papers  (2026-05-31T13:01:18Z)
• LLMQ: Efficient Lower-Precision Pretraining for Consumer GPUs [45.51664355320938]
  We present an end-to-end/C++ implementation for medium-sized language-model training, e.g. 3B to 32B parameters, on commodity GPUs.<n>This is achieved while executing a standard 8-bit training pipeline, without additional algorithmic approximations, and maintaining FLOP utilization of around 50%.
  arXiv  Detail & Related papers  (2025-12-17T10:51:45Z)
• A Scalable Multi-GPU Framework for Encrypted Large-Model Inference [5.966282323502589]
  Encrypted AI using fully homomorphic encryption (FHE) provides strong privacy guarantees.<n>Recent works proposed ASICs to accelerate FHE, but require expensive advanced manufacturing processes that constrain their computation.<n>This paper presents Cerium, a multi- GPU framework for FHE inference on large models.
  arXiv  Detail & Related papers  (2025-12-12T04:15:38Z)
• Minute-Long Videos with Dual Parallelisms [57.22737565366549]
  Diffusion Transformer (DiT)-based video diffusion models generate high-quality videos at scale but incur prohibitive processing latency and memory costs for long videos.<n>We propose a novel distributed inference strategy, termed DualParal.<n>Instead of generating an entire video on a single GPU, we parallelize both temporal frames and model layers across GPUs.
  arXiv  Detail & Related papers  (2025-05-27T11:55:22Z)
• HeadInfer: Memory-Efficient LLM Inference by Head-wise Offloading [79.38548165722229]
  HEADINFER offloads the KV cache to CPU RAM while avoiding the need to fully store the KV cache for any transformer layer on the GPU. We demonstrate HEADINFER maintains computational efficiency while significantly reducing memory footprint.
  arXiv  Detail & Related papers  (2025-02-18T06:26:05Z)
• Efficient LLM Inference with I/O-Aware Partial KV Cache Recomputation [7.204881999658682]
  Inference for Large Language Models (LLMs) is computationally demanding. To reduce the cost of auto-regressive decoding, Key-Value ( KV) caching is used to store intermediate activations. The memory required for KV caching grows rapidly, often exceeding the capacity of GPU memory. A cost-effective alternative is to offload KV cache to CPU memory, which alleviates GPU memory pressure but shifts the bottleneck to the limited bandwidth of the PCIe connection between the CPU and GPU.
  arXiv  Detail & Related papers  (2024-11-26T04:03:14Z)
• MoE-Lightning: High-Throughput MoE Inference on Memory-constrained GPUs [55.95879347182669]
  MoE architecture is renowned for its ability to increase model capacity without a proportional increase in inference cost. MoE-Lightning introduces a novel CPU-GPU-I/O pipelining schedule, CGOPipe, with paged weights to achieve high resource utilization. MoE-Lightning can achieve up to 10.3x higher throughput than state-of-the-art offloading-enabled LLM inference systems for Mixtral 8x7B on a single T4 GPU (16GB)
  arXiv  Detail & Related papers  (2024-11-18T01:06:12Z)
• Practical offloading for fine-tuning LLM on commodity GPU via learned sparse projectors [11.127604539303373]
  Fine-tuning large language models (LLMs) requires significant memory, often exceeding the capacity of a single GPU. A common solution to this memory challenge is offloading compute and data from the GPU to the CPU. We present an offloading framework, LSP-Offload, that enables near-native speed LLM fine-tuning on commodity hardware.
  arXiv  Detail & Related papers  (2024-06-14T16:59:11Z)
• FusionAI: Decentralized Training and Deploying LLMs with Massive Consumer-Level GPUs [57.12856172329322]
  We envision a decentralized system unlocking the potential vast untapped consumer-level GPU. This system faces critical challenges, including limited CPU and GPU memory, low network bandwidth, the variability of peer and device heterogeneity.
  arXiv  Detail & Related papers  (2023-09-03T13:27:56Z)
• Full Parameter Fine-tuning for Large Language Models with Limited Resources [55.794732214059806]
  Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but demand massive GPU resources for training. We propose a new computation, LOw-Memory Optimization (LOMO), which fuses the gradient and the parameter update in one step to reduce memory usage.
  arXiv  Detail & Related papers  (2023-06-16T11:37:15Z)
• FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPU [89.2451963569343]
  FlexGen is a generation engine for running large language model (LLM) inference on a single commodity GPU. When running OPT-175B on a single 16GB GPU, FlexGen achieves significantly higher throughput compared to state-of-the-art offloading systems. On the HELM benchmark, FlexGen can benchmark a 30B model with a 16GB GPU on 7 representative sub-scenarios in 21 hours.
  arXiv  Detail & Related papers  (2023-03-13T05:19:28Z)
• EVEREST: Efficient Masked Video Autoencoder by Removing Redundant Spatiotemporal Tokens [57.354304637367555]
  We present EVEREST, a surprisingly efficient MVA approach for video representation learning. It finds tokens containing rich motion features and discards uninformative ones during both pre-training and fine-tuning. Our method significantly reduces the computation and memory requirements of MVA.
  arXiv  Detail & Related papers  (2022-11-19T09:57:01Z)
• PARIS and ELSA: An Elastic Scheduling Algorithm for Reconfigurable Multi-GPU Inference Servers [0.9854614058492648]
  NVIDIA's Ampere GPU architecture provides features to "reconfigure" one large, monolithic GPU into multiple smaller "GPU partitions" In this paper, we study this emerging GPU architecture with reconfigurability to develop a high-performance multi-GPU ML inference server.
  arXiv  Detail & Related papers  (2022-02-27T23:30:55Z)
• PLSSVM: A (multi-)GPGPU-accelerated Least Squares Support Vector Machine [68.8204255655161]
  Support Vector Machines (SVMs) are widely used in machine learning. However, even modern and optimized implementations do not scale well for large non-trivial dense data sets on cutting-edge hardware. PLSSVM can be used as a drop-in replacement for an LVM.
  arXiv  Detail & Related papers  (2022-02-25T13:24:23Z)
• AxoNN: An asynchronous, message-driven parallel framework for extreme-scale deep learning [1.5301777464637454]
  AxoNN is a parallel deep learning framework that exploits asynchrony and message-driven execution to schedule neural network operations on each GPU. By using the CPU memory as a scratch space for offloading data periodically during training, AxoNN is able to reduce GPU memory consumption by four times.
  arXiv  Detail & Related papers  (2021-10-25T14:43:36Z)
• RTGPU: Real-Time GPU Scheduling of Hard Deadline Parallel Tasks with Fine-Grain Utilization [5.02836935036198]
  We propose RTGPU, which can schedule the execution of multiple GPU applications in real-time to meet hard deadlines. Our approach provides superior schedulability compared with previous work, and gives real-time guarantees to meet hard deadlines for multiple GPU applications.
  arXiv  Detail & Related papers  (2021-01-25T22:34:06Z)
• Nimble: Lightweight and Parallel GPU Task Scheduling for Deep Learning [7.43260596107574]
  We propose Nimble, a deep learning (DL) execution engine that runs tasks in parallel with minimal scheduling overhead. Nable automatically parallelizes the execution of GPU tasks by exploiting multiple GPU streams in a single GPU. evaluation on a variety of neural networks shows that compared to PyTorch, Nimble speeds up inference and training by up to 22.34$times$ and 3.61$times$, respectively.
  arXiv  Detail & Related papers  (2020-12-04T17:25:46Z)

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.