An apparatus and operating method for performing LSTM neural network operation

Abstract

An apparatus for performing LSTM neural network operations and an operating method are provided. The device comprises a direct memory access unit, an instruction cache unit, a controller unit, a plurality of data cache units arranged in parallel and a plurality of data processing modules arranged in parallel, wherein the plurality of data processing modules are in one-to-one correspondence with the data cache units, and are used for obtaining input data from the corresponding data cache units and weights and offsets required during operation, and performing LSTM neural network operation; and parallel operation is executed among the plurality of data processing modules. According to the method, special instructions are adopted for operation, the number of instructions required by operationis greatly reduced, and decoding expenditure is reduced; caching the weight and the bias, so that the data transmission overhead is reduced; the method is not limited to the specific application field, can be used in the fields of voice recognition, text translation, music synthesis and the like, and is high in expandability; and the plurality of data processing modules operate in parallel, so that the operation speed of the LSTM network is remarkably improved.

Description

technical field [0001] The present invention relates to the technical field of neural network operations, and more particularly relates to a device and an operation method for performing LSTM neural network operations. Background technique [0002] Long short-term memory network (LSTM) is a time recurrent neural network (RNN). Due to the unique structural design of the network itself, LSTM is suitable for processing and predicting important events with very long intervals and delays in time series. Compared to traditional recurrent neural networks, LSTM networks show better performance and are well suited for learning from experience to classify, process, and predict time series when there is an unknown amount of time between important events. At present, LSTM networks are widely used in many fields such as speech recognition, video description, machine translation and automatic music synthesis. At the same time, with the continuous deepening of research on LSTM networks, t...

AI Technical Summary

Problems solved by technology

Since the GPU is a device specially used to perform graphics and image operations and scientific computing, it does not provide special support for LSTM networks, and still requires a lot of front-end decoding work to perform LSTM network operations, which will bring a lot of additional overhead

In addition, the GPU has only a small on-chip cache, and the relevant parameters used in the LSTM network need to be repeatedly moved from off-chip, and the off-chip bandwidth has also become a performance bottleneck.

Images