Acceleration device and method for deep learning service

Abstract

The invention discloses an acceleration device for the deep learning service, wherein the device is used for the deep learning calculation on to-be-processed data in a server. The device comprises a network card arranged at a server side, a calculation control module connected with the server through a bus, a first memory and a second memory. The calculation control module is a programmable logic device and comprises a control unit, a data storage unit, a logic storage unit, a bus interface, a first communication interface and a second communication interface, wherein the bus interface, the first communication interface and the second communication interface are respectively communicated with the network card, the first memory and the second memory. The logic storage unit is used for storing the depth learning control logic. The first memory is used for storing the weight data and the offset data of each layer of the network. Based on the device, the calculation efficiency can be effectively improved, and the performance power consumption ratio can also be improved.

Description

technical field [0001] The invention relates to the field of circuit design, in particular to an acceleration device and method for deep learning services. Background technique [0002] With the large-scale successful application of deep learning algorithms in the fields of speech recognition, image recognition, and natural language understanding, the number and frequency of users using deep learning-related services is gradually increasing; Concurrent computing capabilities for related business responses. There are two main methods to improve concurrent computing capabilities: one is to simply increase the CPU to increase concurrent computing capabilities; the other is to use a CPU+GPU heterogeneous system to increase concurrent computing capabilities. Simply increasing the CPU to improve the concurrent computing capability will linearly increase the number of CPUs as the business grows, and the cost will also increase linearly. Moreover, the CPU has limited computing comp...

AI Technical Summary

Problems solved by technology

Simply increasing the CPU to improve the concurrent computing capability will linearly increase the number of CPUs as the business grows, and the cost will also increase linearly. Moreover, the CPU has limited computing components for user programs. If it is used for deep learning computing, Its CPU resources are difficult to fully utilize

Using CPU+GPU heterogeneous system to improve concurrent computing capabilities, because GPU is a processing chip specially designed for graphics computing, if it is used for deep learning computing, it will inevitably lead to problems such as inappropriate and incomplete utilization of some resources

Images