Multitask parallel processing method and apparatus, computer device and storage medium
AssigneeLynxi Technologies Co., Ltd.
InventorsYangshu SHEN, Yaolong ZHU, Wei HE, Luping SHI
Life of this patent
Estimating the term…
Abstract
Embodiments of the present disclosure provide multitask parallel processing method and apparatus, a computer device and a storage medium. The method is applied to a neural network consisting of a plurality of nodes, the neural network including at least one closed-loop path, and the method includes: inputting a data sequence to be computed into the neural network in a form of data packets, each of the data packets including multiple pieces of data; and computing, by the nodes in the closed-loop path, all the data in a currently received data packet each time a computation flow is started.
Claims17
A multitask parallel processing method, which is applied to a neural network consisting of a plurality of nodes, the neural network comprising at least one closed-loop path, and the method comprising:
- inputting a data sequence to be computed into the neural network in a form of data packets, each of the data packets comprising multiple pieces of data;
- and
- computing, by the nodes in the closed-loop path, all the data in a currently received data packet each time a computation flow is started,
- wherein the step of computing, by the nodes in the closed-loop path, all the data in the currently received data packet each time a computation flow is started comprises:
- starting, by the nodes in the closed-loop, a computation flow upon reception of a first piece of data in a data packet, and computing the respective pieces of data in the data packet that are received successively, and
- outputting, by the nodes in the closed-loop path, the computed data each time at least one piece of data in the data packet is computed, wherein the computed at least one piece of data comprises a number of data pieces less than in the data packet.
(canceled)
A multitask parallel processing apparatus, which is applied to a neural network consisting of a plurality of nodes, the neural network comprising at least one closed-loop path, and the apparatus comprising:
- a data input module configured to input a data sequence to be computed into the neural network in a form of data packets, each of the data packets comprising multiple pieces of data;
- and
- a data computation module configured to compute, by the nodes in the closed-loop path, all the data in a currently received data packet each time a computation flow is started,
- wherein the data computation module is configured to:
- start, by the nodes in the closed-loop path, a computation flow upon reception of a first piece of data in a data packet, and compute the respective pieces of data in the data packet that are received successively, and
- output, by the nodes in the closed-loop path, the computed data each time at least one piece of data in the data packet is computed, where the computed at least one piece of data comprises a number of data pieces less than that in the data packet.
Dependencies are inferred from the claim text.