Operator operation scheduling method and device

Abstract

The invention provides an operator operation scheduling method and device. The operator operation scheduling method comprises that: operator parameters and processor parameters corresponding to operator operation are acquired; n scheduling strategies are created according to the operator parameters and the processor parameters, the N scheduling strategies are classified into M scheduling strategysubsets, and each scheduling strategy subset comprises at least one scheduling strategy; the M scheduling strategy subsets are filtered respectively according to the operator parameters and the processor parameters to obtain K feasible scheduling strategies, the K feasible scheduling strategies are optimal scheduling strategies of the K feasible scheduling subsets in the M scheduling strategy subsets, and K is a natural number less than or equal to M; the operator parameters and the K feasible scheduling strategies are input into a cost model to obtain K operator operation costs, wherein the Koperator operation costs are in one-to-one correspondence with the K feasible scheduling strategies; and an optimal scheduling strategy for operator operation is determined according to the target demand and the K operator operation costs. According to the invention, the general processing of the processor architecture and the operator type is realized, and the operator operation performance is improved.

Description

technical field [0001] The present application relates to artificial intelligence technology, and in particular to an operator operation scheduling method and device. Background technique [0002] With the rapid development of artificial intelligence (AI), more and more practical applications begin to use deep learning technology, such as speech recognition, machine translation, automatic driving, etc., deep learning technology in real life applications It has great potential, which makes it more and more attention. The computing efficiency of Deep Neural Network (DNN for short) adopted by deep learning technology directly affects the effect of practical applications, for example, the computing time of algorithms such as target detection, target recognition, and motion prediction in automatic driving. Determines the usability and security of the algorithm. Therefore, how to perform AI computing with high performance is an urgent need for the landing of industrial products....

AI Technical Summary

Problems solved by technology

[0005] However, the above method is only for one kind of operator operation, and its versatility is not high, and it needs to use ANN to traverse all possible operator operation spaces, which is very time-consuming, and it also needs to run the scheduling strategy multiple times on the actual machine, which cannot meet the requirements of the scheduling strategy. millisecond-level real-time output

Images