Apparatus and method for executing recurrent neural network and LSTM computations
AssigneeCambricon (Xi'an) Semiconductor Co., Ltd.
InventorsQi Guo, Xunyu Chen, Yunji Chen, Tianshi Chen
Life of this patent
Estimating the term…
Abstract
Aspects for Long Short-Term Memory (LSTM) blocks in a recurrent neural network (RNN) are described herein. As an example, the aspects may include one or more slave computation modules, an interconnection unit, and a master computation module collectively configured to calculate an activated input gate value, an activated forget gate value, a current cell status of the current computation period, an activated output gate value, and a forward pass result.
Claims28
An apparatus for forward pass in a recurrent neural network (RNN) network, comprising:
- one or more slave computation modules configured to calculate a first input gate partial sum, a second input gate partial sum, and a third input gate partial sum;
- an interconnection unit configured to add the first input gate partial sum, the second input gate partial sum, and the third input gate partial sum to generate a dormant input gate value;
- and
- a master computation module connected to the one or more slave computation modules via the interconnection unit, wherein the master computation module configured to activate the dormmate input gate value to generate an activated input gate value.
A method for forward pass in a recurrent neural network (RNN) network, comprising:
- calculating, by one or more slave computation modules, a first input gate partial sum, a second input gate partial sum, and a third input gate partial sum;
- adding, by an interconnection unit, the first input gate partial sum, the second input gate partial sum, and the third input gate partial sum to generate a dormant input gate value;
- and
- activating, by a master computation module connected to the one or more slave computation modules via the interconnection unit, the dormmate input gate value to generate an activated input gate value.
Dependencies are inferred from the claim text.