CN108399458A - A kind of deep learning model training systems based on SAAS structures - Google Patents
A kind of deep learning model training systems based on SAAS structures Download PDFInfo
- Publication number
- CN108399458A CN108399458A CN201810250383.6A CN201810250383A CN108399458A CN 108399458 A CN108399458 A CN 108399458A CN 201810250383 A CN201810250383 A CN 201810250383A CN 108399458 A CN108399458 A CN 108399458A
- Authority
- CN
- China
- Prior art keywords
- training
- saas
- application
- server
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a kind of deep learning model training systems based on SAAS structures to be based on the method and system that area mechanism builds security system, including：Gateway, training server, application server, distributed file system and database, training server includes training management person and several trainers, SAAS consumer connects Training Server upload and management training collection by gateway, and trained manager distributes to trainer and runs training after checking；Application server includes application manager and several application persons, SAAS consumer connects application server by gateway, distributed by application manager or started application person's operation trained model, later, SAAS consumer is passed to input data by gateway, calls the trained model treatment input data on application person.The present invention does not need consumer and understands many professional knowledges about deep learning；Save expensive hardware cost；Client can be with pay-for-use, easy to use and flexible.
Description
Technical field
The present invention relates to a kind of deep learning model training systems, and in particular to a kind of deep learning based on SAAS structures
Model training systems belong to field of computer technology.
Background technology
The concept of deep learning is derived from the research of artificial neural network, is a new field in machine learning research,
Its motivation is that foundation, simulation human brain carry out the neural network of analytic learning, it imitates the mechanism of human brain to explain data, such as
Image, sound and text.
Many people are intended to the deep learning model of oneself training, and that them can be helped to solve the problems, such as is finer, but this
Process may require that consumer understands many professional knowledges about deep learning and expensive hardware supported, greatly hinder depth
The universal and application of study.
Therefore, a kind of deep learning model training systems based on SAAS (software is service) structure are developed, provide one
The SAAS of kind the deep learning model for training customized training system and the pay-for-use about deep learning model
System is not only very important, and also has important application prospect.
Invention content
The present invention has made improvements in view of the above-mentioned problems of the prior art, discloses a kind of depth built based on SAAS
Model training systems are practised, are provided a kind of for training the deep learning model of customized training system and about depth
Practise the SAAS systems of the pay-for-use of model.
In order to realize that above-mentioned target, the technical solution adopted in the present invention be：
A kind of deep learning model training systems based on SAAS structures, including：Gateway (1), is answered at training server (2)
With server (3), distributed file system (4) and database (5),
The training server (2) includes training management person (21) and several trainers (22), and SAAS consumer passes through
The gateway (2) connects Training Server (2) upload and management training collection, divides after the training management person (21) checks
Trainer described in dispensing (22) runs training, stores trained model after the completion；
The application server (3) includes application manager (31) and several application persons (32), and SAAS consumer passes through
The gateway (2) connects the application server (3), is distributed by the application manager (31) or started an application person
(32) trained model is run, later, SAAS consumer is passed to described in input data, calling by the gateway (2) and is answered
Trained model treatment input data on user (32), after the completion return result；
The training set of distributed file system (4) storage SAAS consumer's upload, trained model；The number
Metadata is stored and managed according to library (5).
Deep learning model training systems above-mentioned based on SAAS structures, the application manager (31) distribution, start,
The application that supervision and stopping are left unused passes through the training pattern in the application person (32) unloading application.
It is above-mentioned based on SAAS structure deep learning model training systems, the training server (2) be server or
Server cluster.
It is above-mentioned based on SAAS structure deep learning model training systems, the application server (3) be server or
Server cluster.
It is above-mentioned based on SAAS structure deep learning model training systems, the gateway (1) be include load balancing work(
It can be with the gateway of billing function.
It is above-mentioned based on SAAS structure deep learning model training systems, the distributed file system (4) include but
It is not limited to following type file system：NFS、AFS、KFS、DFS.
Compared with prior art, the invention has the beneficial effects that：
(1) it does not need consumer and understands many professional knowledges about deep learning；
(2) expensive hardware cost is saved；
(3) client can be with pay-for-use, easy to use and flexible.
Description of the drawings
Fig. 1 is the system of a specific embodiment of the deep learning model training systems of the present invention built based on SAAS
Structural schematic diagram；
Fig. 2 is the model instruction of a specific embodiment based on the SAAS deep learning model training systems built in Fig. 1
Practice flow chart；
Fig. 3 is that the model of a specific embodiment based on the SAAS deep learning model training systems built in Fig. 1 is answered
Use flow chart.
Wherein：
1- gateway 2- training servers
21- training management person 22- trainers
3- application server 31- application managers
32- application person's 4- distributed file systems
5- databases.
Specific implementation mode
Specific introduce is made to the present invention below in conjunction with the drawings and specific embodiments.
Referring to FIG. 1, FIG. 2 and FIG. 3, the deep learning model training systems of the invention based on SAAS structures, including：Gateway
1, training server 2, application server 3, distributed file system 4 and database 5,
Training server 2 includes training management person 21 and several trainers 22, and SAAS consumer is connected by gateway 2 and trained
The upload of server 2 and management training collection are instructed, trained manager 21 distributes to trainer 22 and runs training, deposits after the completion after checking
Store up trained model；
Training set is one group of data for being used for training machine learning model, having divided class (accomplishing fluently label), for example, instruction
Practice the machine learning model for distinguishing cat and dog photo, the photo of the photo and one group of dog that need one group of cat is used as training set；Compare again
Such as, training, which is distinguished, comments the still poor machine learning model commented in favorable comment, the word that this three classes is commented on is needed to be used as training set.
Application server 3 includes application manager 31 and several application persons 32, and SAAS consumer passes through the gateway (2)
The application server (3) is connected, distributed by the application manager (31) or starts application person (32) operation
Trained model, later, SAAS consumer are passed to input data by the gateway (2), call on the application person (32)
Trained model treatment input data, return result after the completion；
The training set of the storage SAAS consumer's upload of distributed file system 4, trained model；Database 5 stores
With management metadata；Metadata is the data of data, refers to the summary info of training set and model in the present system, confidence is matched in training
Iterations, the learning rate of classify, each the classify sample data volume, training pattern that include of breath etc., model name, training set
Deng.
As a preferred solution, the idle application of the distribution of application manager 31, startup, supervision and stopping, by answering
User 32 unloads the training pattern in application.
As a preferred solution, training server 2 is server or server cluster.
As a preferred solution, application server 3 is server or server cluster.
As a preferred solution, gateway 1 is the gateway for including load-balancing function and billing function.
As a preferred solution, distributed file system 4 includes but not limited to following type file system：NFS、
AFS, KFS, DFS, preferably NFS.
It should be noted that the foregoing is merely presently preferred embodiments of the present invention, it is not intended to limit the invention, it is all at this
Within the spirit and principle of invention, any modification, equivalent replacement, improvement and so on should be included in the protection model of the present invention
Within enclosing.
Claims (6)
1. a kind of deep learning model training systems based on SAAS structures, which is characterized in that including：Gateway (1), training service
Device (2), application server (3), distributed file system (4) and database (5),
The training server (2) includes training management person (21) and several trainers (22), and SAAS consumer passes through described
Gateway (2) connects Training Server (2) upload and management training collection, is distributed to after the training management person (21) checks
Trainer (22) the operation training, stores trained model after the completion；
The application server (3) includes application manager (31) and several application persons (32), and SAAS consumer passes through described
Gateway (2) connects the application server (3), is distributed by the application manager (31) or started an application person (32)
Trained model is run, later, SAAS consumer passes through the gateway (2) and is passed to input data, calls the application person
(32) the trained model treatment input data on, after the completion return result；
The training set of distributed file system (4) storage SAAS consumer's upload, trained model；The database
(5) store and manage metadata.
2. a kind of deep learning model training systems based on SAAS structures according to claim 1, which is characterized in that institute
It states application manager (31) distribution, start, the application that supervision and stopping are idle, unloaded in application by the application person (32)
Training pattern.
3. a kind of deep learning model training systems based on SAAS structures according to claim 1, which is characterized in that institute
It is server or server cluster to state training server (2).
4. a kind of deep learning model training systems based on SAAS structures according to claim 1, which is characterized in that institute
It is server or server cluster to state application server (3).
5. a kind of deep learning model training systems based on SAAS structures according to claim 1, which is characterized in that institute
It is the gateway for including load-balancing function and billing function to state gateway (1).
6. a kind of deep learning model training systems based on SAAS structures according to claim 1, which is characterized in that institute
It includes but not limited to following type file system to state distributed file system (4)：NFS、AFS、KFS、DFS.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810250383.6A CN108399458B (en) | 2018-03-26 | 2018-03-26 | Deep learning model training system constructed based on SAAS |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810250383.6A CN108399458B (en) | 2018-03-26 | 2018-03-26 | Deep learning model training system constructed based on SAAS |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108399458A true CN108399458A (en) | 2018-08-14 |
| CN108399458B CN108399458B (en) | 2022-07-08 |
Family
ID=63093141
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810250383.6A Active CN108399458B (en) | 2018-03-26 | 2018-03-26 | Deep learning model training system constructed based on SAAS |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108399458B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241141A (en) * | 2018-09-04 | 2019-01-18 | 北京百度网讯科技有限公司 | The training data treating method and apparatus of deep learning |
| CN109460826A (en) * | 2018-10-31 | 2019-03-12 | 北京字节跳动网络技术有限公司 | For distributing the method, apparatus and model modification system of data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598239A (en) * | 2015-01-19 | 2015-05-06 | 中国传媒大学 | Software issuing and using system, and software using and issuing method based on system |
| US20160125039A1 (en) * | 2014-10-30 | 2016-05-05 | Szegedi Tudományegyetem | Data mining method and apparatus, and computer program product for carrying out the method |
| CN106529673A (en) * | 2016-11-17 | 2017-03-22 | 北京百度网讯科技有限公司 | Deep learning network training method and device based on artificial intelligence |
| CN107370796A (en) * | 2017-06-30 | 2017-11-21 | 香港红鸟科技股份有限公司 | A kind of intelligent learning system based on Hyper TF |
| CN107563417A (en) * | 2017-08-18 | 2018-01-09 | 北京天元创新科技有限公司 | A kind of deep learning artificial intelligence model method for building up and system |
-
2018
- 2018-03-26 CN CN201810250383.6A patent/CN108399458B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160125039A1 (en) * | 2014-10-30 | 2016-05-05 | Szegedi Tudományegyetem | Data mining method and apparatus, and computer program product for carrying out the method |
| CN104598239A (en) * | 2015-01-19 | 2015-05-06 | 中国传媒大学 | Software issuing and using system, and software using and issuing method based on system |
| CN106529673A (en) * | 2016-11-17 | 2017-03-22 | 北京百度网讯科技有限公司 | Deep learning network training method and device based on artificial intelligence |
| CN107370796A (en) * | 2017-06-30 | 2017-11-21 | 香港红鸟科技股份有限公司 | A kind of intelligent learning system based on Hyper TF |
| CN107563417A (en) * | 2017-08-18 | 2018-01-09 | 北京天元创新科技有限公司 | A kind of deep learning artificial intelligence model method for building up and system |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241141A (en) * | 2018-09-04 | 2019-01-18 | 北京百度网讯科技有限公司 | The training data treating method and apparatus of deep learning |
| CN109241141B (en) * | 2018-09-04 | 2022-04-29 | 北京百度网讯科技有限公司 | Deep learning training data processing method and device |
| CN109460826A (en) * | 2018-10-31 | 2019-03-12 | 北京字节跳动网络技术有限公司 | For distributing the method, apparatus and model modification system of data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108399458B (en) | 2022-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mumby | Theorizing resistance in organization studies: A dialectical approach | |
| Wightman et al. | Part-task training for tracking and manual control | |
| DE102013212215A1 (en) | System and method for determining expertise through speech analysis | |
| CN109002515A (en) | A kind of method and apparatus of intelligent response | |
| CN108399458A (en) | A kind of deep learning model training systems based on SAAS structures | |
| Pinet et al. | “Twisting fingers”: The case for interactivity in typed language production | |
| Winters | Digital history | |
| Hamidi Rad et al. | Pytfl: A python-based neural team formation toolkit | |
| Jeners et al. | Towards an integration of multiple process improvement reference models based on automated concept extraction | |
| Muran et al. | A brief introduction to psychotherapy research. | |
| Bulla et al. | Comparing user perspectives in a virtual reality cultural heritage environment | |
| Koike et al. | Learner model for adaptive scaffolding in intelligent tutoring systems for organizing programming knowledge | |
| Martinez-Maldonado et al. | Modelling embodied mobility teamwork strategies in a simulation-based healthcare classroom | |
| Bernius | Toward Computer-Aided Assessment of Textual Exercises in Very Large Courses | |
| Gomez et al. | A framework to support interoperable Game‐based Assessments as a Service (GBAaaS): Design, development, and use cases | |
| Suciu et al. | Generating hints for programming problems without a solution | |
| Esabella et al. | Development of application information system in rhee district | |
| Wu et al. | ILLINOISCLOUDNLP: Text Analytics Services in the Cloud. | |
| Lőrincz et al. | Development of an expert system framework for lifestyle improvement | |
| Reel et al. | Dispatch from the ivory tower: Reflections on the 20th anniversary of Deconstructing Developmental Psychology | |
| Hallberg | Using Low-Code Platforms to Collect Patient-Generated Health Data: A Software Developer’s Perspective | |
| Sawamoto et al. | Extraction of important interactions in medical interviewsusing nonverbal information | |
| Pauli et al. | Achieving a qualitative reproduction of polychronization: computation with spikes | |
| Terhorst et al. | Introduction to the simulation of structurally detailed large-scale neuronal networks (NEST) | |
| Klijn et al. | NestMC: A morphologically detailed neural network simulator for modern high performance computer architectures |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |