PICO: Pipeline Inference Framework for Versatile CNNs on Diverse Mobile Devices

TL;DR

PICO is a framework that optimizes CNN inference across multiple mobile devices by intelligently partitioning models and mapping workloads, significantly improving throughput in heterogeneous environments.

Contribution

Introduces a generic graph partition and many-to-many mapping algorithm for efficient CNN inference on diverse mobile devices.

Findings

Achieves 1.8 to 6.8 times throughput improvement

Effectively handles heterogeneous devices and wireless communication

Supports versatile CNN architectures

Abstract

Distributing the inference of convolutional neural network (CNN) to multiple mobile devices has been studied in recent years to achieve real-time inference without losing accuracy. However, how to map CNN to devices remains a challenge. On the one hand, scheduling the workload of state-of-the-art CNNs with multiple devices is NP-Hard because the structures of CNNs are directed acyclic graphs (DAG) rather than simple chains. On the other hand, distributing the inference workload suffers from expensive communication and unbalanced computation due to the wireless environment and heterogeneous devices. This paper presents PICO, a pipeline cooperation framework to accelerate the inference of versatile CNNs on diverse mobile devices. At its core, PICO features: (1) a generic graph partition algorithm that considers the characteristics of any given CNN and orchestrates it into a list of model pieces with suitable granularity, and (2) a many-to-many mapping algorithm that produces the best pipeline configuration for heterogeneous devices. In our experiment with 2 ~ 8 Raspberry-Pi devices, the throughput can be improved by 1.8 ~ 6.8x under different CPU frequencies.