Title:
NEURAL NETWORK SCHEDULING METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2021/237755
Kind Code:
A1
Abstract:
A neural network scheduling method and apparatus, which relate to the field of neural networks, can increase the utilization rate of the storage capacity on a chip, and improve the operating performance of hardware. The method comprises: determining a first batch size respectively corresponding to each layer in a neural network (S701); on the basis of the first batch size, segmenting the neural network into a neural network that contains at least one first layer group (S702); on the basis of the segmentation result of the first layer group, segmenting the neural network into a neural network that contains at least one second layer group (S703); and on the basis of the segmentation result of the second layer group, scheduling the neural network (S704).
Inventors:
YUAN HONGHUI (CN)
LI SHUCHENG (CN)
XIONG LEJIN (CN)
NIKITA CHERNEGA (CN)
LI SHUCHENG (CN)
XIONG LEJIN (CN)
NIKITA CHERNEGA (CN)
Application Number:
PCT/CN2020/093544
Publication Date:
December 02, 2021
Filing Date:
May 29, 2020
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/063
Foreign References:
| CN107454965A | 2017-12-08 | |||
| CN110058943A | 2019-07-26 | |||
| CN110447044A | 2019-11-12 | |||
| US20190303762A1 | 2019-10-03 |
Other References:
See also references of EP 4148627A4
Attorney, Agent or Firm:
BEIJING ZBSD PATENT&TRADEMARK AGENT LTD. (CN)
Download PDF:
Previous Patent: ROAD CONDITION DETECTION METHOD AND APPARATUS
Next Patent: BUILDING BLOCK FIGURE TOY
Next Patent: BUILDING BLOCK FIGURE TOY