Unlock instant, AI-driven research and patent intelligence for your innovation.

A method for predicting layer computation time in deep learning models using similar layers

A technology of computing time and prediction layer, applied in the field of deep learning, can solve long-time problems, achieve low overhead and avoid repeated measurement

Active Publication Date: 2021-05-11
CLUSTAR TECH LO LTD
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] It is precisely because of the above reasons that it obviously takes longer to train a deep learning model with only a single node; in this regard, the current mainstream deep learning frameworks support the realization of distributed training models through parallel computing

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A method for predicting layer computation time in deep learning models using similar layers

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0063] The technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Apparently, the described embodiments are only part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0064] The following is a preferred embodiment of the present invention.

[0065] figure 1 A schematic flowchart of a method for predicting the calculation time of layers in a deep learning model by using similar layers provided in this embodiment. The above embodiments demonstrate the prediction of the layer calculation time of the model implemented under the framework of tensorflow.

[0066] like figure 1 as shown,

[0067] 1) Before prediction, the la...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The present invention provides a method for predicting the calculation time of layers in a deep learning model by using similar layers. This method evaluates similar layers through the key attributes and non-key attributes of layers, evaluates the similarity between similar layers through the partial order relationship of non-key attributes of layers, and then uses the computing time of similar layers in the historical log to estimate the computing time of the layer to be predicted ; For those that cannot be estimated (that is, there is no similar layer for estimation in the historical log) or the estimation is not acceptable, measure the layer. The present invention avoids some repetitive measurement work and reduces system overhead by reusing the known layer calculation time. In addition, the present invention also realizes the localization of prediction, and the prediction of the calculation time of the layer of the model can be completed on a local device.

Description

technical field [0001] The invention relates to the technical field of deep learning; in particular, it relates to a method for predicting the calculation time of layers in a deep learning model by using similar layers. Background technique [0002] Deep learning has been widely used in industries such as finance and insurance, security monitoring and so on. The core method of deep learning is to use neural network models to analyze and describe the characteristics of data. Training a deep learning model requires multiple rounds of iterations on the order of seconds. Secondly, compared with linear algorithms, deep learning often requires a larger amount of data training to obtain an accurate model. Training a neural network model often takes days or weeks to complete. [0003] It is precisely because of the above reasons that it will obviously take longer to train a deep learning model with only a single node; in this regard, the current mainstream deep learning framework...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06N3/04G06N20/00
Inventor 孙军欢张骏雪
Owner CLUSTAR TECH LO LTD