Neural network compiling method for storage and calculation integrated platform

Abstract

The invention discloses a neural network compiling method for a storage and calculation integrated platform, which relates to the field of storage and calculation integration, and comprises the following steps: analyzing a neural network model, and mapping the neural network model into intermediate representation described by calculation nodes; optimizing a calculation graph; converting into operator-level intermediate representation; carrying out operator task division and binding with a hardware basic unit; performing operator-level optimization, and reducing the number of times of reading discontinuous memories and the number of times of weight mapping. According to the invention, the calculation flow graph and the neural network operator are optimized according to the characteristics of storage and calculation integrated calculation, the expenditure of writing back an intermediate result between graph-level operators is reduced, and the frequency of remapping the weight when the storage and calculation resources are insufficient is reduced.

Description

technical field [0001] The invention relates to the field of storage-computing integration, in particular to a neural network compiling method for a storage-computing integration platform. Background technique [0002] Deep learning has made many breakthroughs in speech recognition, image recognition and other fields. The existing deep neural network needs to complete calculations in a shorter time and with lower power consumption, which puts forward higher requirements for deep learning computing chips. Therefore, an accelerator with integrated storage and calculation using a non-volatile memory (Non-volatile memory, NVM) such as a memristor as a calculation unit appears. This kind of accelerator effectively solves the bottleneck of bandwidth, and has the characteristics of low power consumption and high speed. Its research and development also opens up a new field for in-memory computing. [0003] At present, the basic algorithms of artificial intelligence are relatively ...

AI Technical Summary

Problems solved by technology

[0013] 1. It cannot support a variety of neural network programming frameworks, and has not done more explorations in computational graph-level optimization;

[0014] 2. There is no flexible operator optimization and scheduling interface, and a similar mapping method is adopted for all operators. Therefore, for a new specific operator, it is difficult for programmers to maximize the execution efficiency of the hardware;

[0015] 3. Deploy the weights of the entire network on the array at one time, without considering the need to update the weights. In fact, it is difficult to map the weights of the network to the Crossbararray (cross array) at one time due to the limitation of the technology level and the weight scale of the neural network model.

Images