# LOIP:Collaborative Lossless LLM Inference Serving with Offloading-based Pipeline Parallelism on Edge Devices

arXiv: https://arxiv.org/abs/2512.21835 · HTML: https://mlsanity.com/paper/2512.21835/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Serving systems and request scheduling, one of 12 directions in Efficient LLM inference and serving.
- Builds on 19 of its 43 references that are in the map.
- First posted on arXiv 26 Dec 2025.

## Placed in

- Efficient LLM inference and serving / [Serving systems and request scheduling](https://mlsanity.com/inference/serving-systems-and-request-scheduling/)

## Builds on (most cited first, 19 of 19)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration](https://mlsanity.com/paper/2306.00978/) (2023)
- [SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot](https://mlsanity.com/paper/2301.00774/) (2023)
- [Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes](https://mlsanity.com/paper/2305.02301/) (2023)
- [LLM-Pruner: On the Structural Pruning of Large Language Models](https://mlsanity.com/paper/2305.11627/) (2023)
- [KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization](https://mlsanity.com/paper/2401.18079/) (2024)
- [FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPU](https://mlsanity.com/paper/2303.06865/) (2023)
- [DeepSpeed Inference: Enabling Efficient Inference of Transformer Models at Unprecedented Scale](https://mlsanity.com/paper/2207.00032/) (2022)
- [SliceGPT: Compress Large Language Models by Deleting Rows and Columns](https://mlsanity.com/paper/2401.15024/) (2024)
- [SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression](https://mlsanity.com/paper/2306.03078/) (2023)
- [Less is More: Task-aware Layer-wise Distillation for Language Model Compression](https://mlsanity.com/paper/2210.01351/) (2022)
- [CoEdge: Cooperative DNN Inference with Adaptive Workload Partitioning over Heterogeneous Edge Devices](https://mlsanity.com/paper/2012.03257/) (2020)
- [Galaxy: A Resource-Efficient Collaborative Edge AI System for In-situ Transformer Inference](https://mlsanity.com/paper/2405.17245/) (2024)
- [Jupiter: Fast and Resource-Efficient Collaborative Inference of Generative LLMs on Edge Devices](https://mlsanity.com/paper/2504.08242/) (2025)
- [Unity is Power: Semi-Asynchronous Collaborative Training of Large-Scale Models with Structured Pruning in Resource-Limited Clients](https://mlsanity.com/paper/2410.08457/) (2024)

## Abstract (arXiv)

Providing lossless inference services of LLMs on edge devices remains challenging, especially given the extremely tight memory budgets. The existing offloading techniques inevitably introduce numerous loading bubbles, which further inflate the end-to-end latency of the entire inference pipeline. Meanwhile, dynamically fluctuating network bandwidth and diverse user request patterns pose additional obstacles to efficient lossless inference on edge devices. To address this, we propose LOIP, a collaborative lossless LLM inference system that employs an offloading-based interleaved pipeline parallelism to better overlap model offloading with computing and communicating. Specifically, LOIP first constructs an offloading-aware cost model to characterize inference latency and memory overhead under heterogeneous device capabilities and limited bandwidth. Based on this cost model, LOIP develops a fine-grained allocation scheduler that determines latency-efficient layer partitions across devices while explicitly accounting for offloading overhead, along with a unified memory architecture (UMA)-aware loading optimization using customized CUDA operators to reduce runtime loading overhead. LOIP further designs an online memory adaptation strategy to handle the increasing KV cache pressure and dynamic bandwidth fluctuations during inference. We implement LOIP with 2500+ lines of Python and 500+ lines of C++/CUDA code, and deploy it on five heterogeneous NVIDIA Jetson edge devices for lossless collaborative inference of LLaMA3.3-70B-Instruct. Extensive experiments demonstrate that LOIP achieves 8.8×~20.3× speedups over the SOTA baselines under different bandwidth conditions and request patterns without compromising model accuracy.