Hardware acceleration system for LSTM (Long Short Term Memory) network model

Abstract

The invention discloses a hardware acceleration system for an LSTM network model, and belongs to the technical field of deep learning hardware acceleration. The invention discloses a hardware acceleration system for a deep learning long short-term memory (LSTM) network model. The hardware acceleration system comprises a network reasoning calculation core and a network data storage core. The network reasoning calculation core is used as a calculation accelerator of an LSTM network model, calculation units are deployed according to the network model, and calculation acceleration of the calculation units such as convolution operation, matrix dot multiplication, matrix addition and an activation function is achieved; and the network data storage core serves as a data cache and interaction controller of an LSTM network model, deploys an on-chip cache unit according to the network model, and realizes a data interaction link between the computing core and an off-chip memory. According to the invention, the calculation parallelism of the LSTM network model is improved, the processing delay is reduced, the memory access time is shortened, and the memory access efficiency is improved.

Description

technical field [0001] The invention belongs to the field of deep learning hardware acceleration, and more specifically relates to a hardware acceleration system oriented to an LSTM network model. Background technique [0002] Long Short-Term Memory (LSTM), as a variant of deep learning Recurrent Neural Network (RNN), is widely used in sequence model processing tasks such as speech recognition, natural language processing, and image compression. LSTM effectively solves the problem of gradient explosion and gradient disappearance in the RNN training process by introducing a gating mechanism and a state value for storing long-term and short-term historical information, and relatively greatly increases its computational complexity and space complexity. Its intensive calculation and memory access limit its application on the embedded hardware platform with limited resources. Therefore, it is a very meaningful research to design and accelerate the optimization of the LSTM model f...

AI Technical Summary

Problems solved by technology

Although this method realizes the acceleration of the LSTM model system, it mainly optimizes the bottleneck of storage resource consumption, and the optimization effect at the computing level is average.

[0006] In summary, although the research on LSTM model acceleration tasks for hardware platforms is of great significance, the research results in this area are still relatively scarce, and an LSTM network acceleration design with good versatility and excellent acceleration effects is urgently needed

Images