CN111814911A - Electric power AI training platform based on containerization management and training method thereof - Google Patents
Electric power AI training platform based on containerization management and training method thereof Download PDFInfo
- Publication number
- CN111814911A CN111814911A CN202010823710.XA CN202010823710A CN111814911A CN 111814911 A CN111814911 A CN 111814911A CN 202010823710 A CN202010823710 A CN 202010823710A CN 111814911 A CN111814911 A CN 111814911A
- Authority
- CN
- China
- Prior art keywords
- training
- model
- layer
- electric power
- algorithm
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/40—Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/546—Message passing systems or structures, e.g. queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
- G06V10/95—Hardware or software architectures specially adapted for image or video understanding structured as a network, e.g. client-server architectures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
- G06V10/955—Hardware or software architectures specially adapted for image or video understanding using specific electronic processors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5011—Pool
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/508—Monitor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/54—Indexing scheme relating to G06F9/54
- G06F2209/548—Queue
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an electric power AI training platform based on containerization management and a training method thereof, belonging to the technical field of artificial intelligence, and particularly relating to an electric power AI training platform based on containerization management and a training method thereof, wherein the electric power AI training platform comprises a hardware resource layer, an operating system layer, a communication layer, a database layer, an algorithm layer and a task layer, the hardware resource layer comprises a GPU resource pool and a storage resource pool, the communication layer adopts RabbitMQ multi-language communication, the electric power AI training platform based on containerization management and the training method thereof are based on containerization technology, the customization of the electric power AI training platform is realized by a Web interface access mode, the platform can meet the requirements of various service scenes, the rapid training and the application deployment of the algorithm under the electric power scene are realized by utilizing strong hardware resource configuration, the optimization process and the performance index of the model can be visually displayed, the algorithm model is effectively managed, an iterative training trigger mechanism, dynamic updating of the in-situ algorithm is supported.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an electric power AI training platform based on containerization management and a training method thereof.
Background
In recent years, with the progress of computer technology, machine learning, especially deep learning, has been greatly developed, and still becomes the current most fiery research field, and is widely popularized and applied in many other fields including the power industry.
The research of deep learning needs a larger data set, higher calculation amount and longer operation time, and at this time, the resources (CPU, GPU, memory, disk, etc.) and performance of a single computer easily get into a bottleneck and cannot meet the execution of a deep learning task.
The containerization technology represented by Docker is mature day by day, a virtualized operating environment is created by using a mirror image, the operating environment contains all required dependencies, and the characteristics of light weight and easiness in management are widely welcomed, so that the Docker is used for deploying related components of an AI training platform to form a final AI training platform in a combined mode, a lot of work can be reduced, and meanwhile great support is provided for the robustness of the platform. The invention realizes the customization of the electric power AI training platform by a Web interface access mode on the basis of a containerization technology.
At present, an AI training platform on the market can only use a built-in model of the platform to train and predict, cannot enable a user to develop an algorithm by himself, cannot support customizable model training and prediction, and cannot meet the requirements of various service scenes. Meanwhile, deployment application of the model is bound with a product of the platform, flexible use cannot be achieved, training model management is lacked, model optimization processes and performance indexes cannot be visually displayed, and an iterative training triggering mechanism is lacked, so that the rapid iterative model cannot be achieved, and dynamic updating of a field algorithm is supported.
Disclosure of Invention
The invention aims to provide an electric power AI training platform based on containerization management and a training method thereof, aiming at solving the problems that the existing AI training platform on the market in the background art can only use a built-in model of the platform to train and predict, cannot enable a user to develop an algorithm by himself, cannot support customizable model training and prediction, and cannot meet the requirements of various service scenes. Meanwhile, deployment application of the model is bound with a product of the platform, flexible use cannot be achieved, training model management is lacked, model optimization processes and performance indexes cannot be visually displayed, and an iterative training triggering mechanism is lacked, so that the problems that the model is difficult to iterate rapidly and dynamic updating of a field algorithm is supported are solved.
In order to achieve the purpose, the invention provides the following technical scheme: an electric power AI training platform based on containerization management comprises a hardware resource layer, an operating system layer, a communication layer, a database layer, an algorithm layer and a task layer, wherein the hardware resource layer comprises a GPU resource pool and a storage resource pool, the communication layer adopts RabbitMQ multi-language communication, the database layer designs a corresponding database table structure and an incidence relation thereof according to the training platform algorithm layer and the task layer, the algorithm layer is based on deep learning frames such as Tensorflow, Pythrch, Caffe and the like, a plurality of algorithm model mirror image files are preset, the task layer comprises three deep learning fields of image classification, target detection and image segmentation, the database layer has data set management and data calibration functions, has uploading and downloading functions, and supports data enhancement mode selection and multi-enhancement mode image comparison display.
The training method of the electric power AI training platform based on containerization management comprises the following steps.
S1: creating a project, selecting a training data set, and selecting a built-in algorithm model;
s2: algorithm training, namely selecting a data set label to be trained, the number of GPUs, uploading an algorithm configuration file, and starting a training script;
s3: monitoring the state, namely monitoring the training process and displaying the current training batch and the Loss curve;
s4: after training is finished, storing the optimal model, and displaying a final Loss curve;
s5: and (4) performing algorithm verification, displaying the effect of the model on the verification set, supporting simultaneous verification of multiple models, comparing and displaying the model effect, and displaying the result pictures in multiple columns.
Preferably, the method has a multi-model management function, version information of the model, a test report of the model, and a trigger mechanism for uploading and downloading the model and iterative training.
Preferably, the method comprises the overall evaluation of the model on the verification set, and performance indexes such as a missing report rate, a false report rate and a mAP value, including an overall Loss curve and a correct rate curve of the model.
Preferably, the model is uploaded and downloaded, and the model is downloaded locally and deployed to a field application platform in a supporting mode.
Preferably, after the field data of the iterative training triggering mechanism is newly increased to a set value, the model can be iteratively trained for a set number of times on the basis of the original model, so that the field detection model is optimal, and the rapid deployment of the model is ensured.
Compared with the prior art, the invention has the beneficial effects that: the electric power AI training platform based on containerization management and the training method thereof are based on containerization technology, and the customization of the electric power AI training platform is realized through a Web interface access mode, the platform can meet the requirements of various service scenes, the rapid training and application deployment of the algorithm under the electric power scene are realized by utilizing strong hardware resource allocation, the model optimization process and performance indexes can be visually displayed, the algorithm model is effectively managed, an iterative training trigger mechanism is provided, the rapid iterative model can be realized, and the dynamic update of the field algorithm is supported.
Drawings
FIG. 1 is a schematic diagram of the system design of the present invention;
FIG. 2 is a diagram of a hardware resource architecture according to the present invention;
FIG. 3 is a diagram of an algorithm container management architecture of the present invention;
FIG. 4 is a task creation flow diagram according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides an electric power AI training platform based on containerization management and a training method thereof, which can help the intelligent construction of an electric power system and help a user to quickly train and deploy an AI algorithm model under an electric power scene, and please refer to fig. 1-4, wherein the electric power AI training platform comprises a hardware resource layer, an operating system layer, a communication layer, a database layer, an algorithm layer and a task layer;
referring to fig. 1 again, the modules are electrically connected in series, specifically, the hardware resource layer includes a GPU resource pool and a storage resource pool, the communication layer adopts RabbitMQ multi-language communication, the database layer designs a corresponding database table structure and an association relation thereof according to a training platform algorithm layer and a task layer, the algorithm layer presets a plurality of algorithm model mirror images based on a deep learning framework such as tensflow, pitorch, Caffe and the like, the task layer includes three deep learning fields of image classification, target detection and image segmentation, the database layer has data set management and data calibration functions, has upload and download functions, and supports data enhancement mode selection and multi-enhancement mode image contrast display;
a control method of an electric power AI training platform based on containerization management comprises the following steps:
s1: creating a project, selecting a training data set, and selecting a built-in algorithm model;
s2: algorithm training, namely selecting a data set label to be trained, the number of GPUs, uploading an algorithm configuration file, and starting a training script;
s3: monitoring the state, namely monitoring the training process and displaying the current training batch and the Loss curve;
s4: after training is finished, storing the optimal model, and displaying a final Loss curve;
s5: and (4) performing algorithm verification, displaying the effect of the model on the verification set, supporting simultaneous verification of multiple models, comparing and displaying the model effect, and displaying the result pictures in multiple columns.
Referring to fig. 1 again, in order to support the uploading and downloading of the model and the iterative training triggering mechanism, the method specifically has a multi-model management function, version information of the model, and a test report of the model.
Referring to fig. 1 again, in order to perform the overall evaluation, the method specifically includes the overall evaluation of the model on the verification set, and performance indexes such as a missing report rate, a false report rate, and a mep value, including an overall Loss curve and a correct rate curve of the model.
Referring again to fig. 1, in order to support the uploading and downloading of the model, specifically, the uploading and downloading of the model, the downloading of the model to the local and the deployment to the field application platform are supported.
Referring to fig. 1 again, in order to iteratively train the set number of times on the basis of the original model, specifically, after the field data of the iterative training triggering mechanism is newly increased to the set value, the model may iteratively train the set number of times on the basis of the original model, so as to ensure that the field detection model is optimal and the model is rapidly deployed.
Example 1
An electric power AI training platform based on containerization management comprises a hardware resource layer, an operating system layer, a communication layer, a database layer, an algorithm layer and a task layer, and has the characteristics of openness, sharing, expandability and the like, wherein the hardware resource layer comprises a GPU resource pool and a storage resource pool, the GPU resource pool adopts a mode of a multi-machine multi-card Tesla P40 GPU server cluster to provide hardware resource service for algorithm training, supports the unified management of a mixed cluster consisting of a plurality of hardware, distributes resources by taking GPU cards as granularity, supports the state monitoring management of hardware resources to support load balancing, smoothly expands the capacity and realizes the reasonable scheduling of the hardware resources, the storage resource pool adopts 3 cluster storage servers, has high performance and can realize 4.5 ten thousand pieces of data interaction per second for 35kB pictures; the method supports the scalability, supports hundreds of clusters to the maximum, and linearly expands the concurrency capability; the multi-algorithm unified management and task containerization heterogeneous resource unified scheduling method has the advantages that the availability is high, the duplicate copying of metadata and user data is supported, the stability and the reliability of a platform are guaranteed, the communication layer adopts RabbitMQ multi-language communication, the database layer designs a corresponding database table structure and the incidence relation thereof according to the training platform algorithm layer and the task layer, data support is provided for the training platform, the algorithm layer is based on deep learning frames such as Tensorflow, Pythrch and Caffe, various algorithm model mirror files are preset, the method is suitable for algorithm training under various power scenes, the task layer comprises three deep learning fields of image classification, target detection and image segmentation, the multi-algorithm unified management and the task containerization heterogeneous resource unified scheduling are completed, the database layer has the functions of data set management and data calibration, the uploading and downloading functions are realized, and the data enhancement mode selection and the multi-enhancement mode.
Example 2
The system design scheme of the invention is shown in figure 1 and comprises a hardware resource layer, an operating system layer, a communication layer, a database layer, an algorithm layer and a task layer.
Wherein, the hardware resource layer is as shown in fig. 2: the system comprises a GPU resource pool and a storage resource pool, wherein a Kubernets technology is adopted in a hardware resource layer, and the containerization application on a plurality of hosts is simple and efficient in a container deployment mode.
Specifically, the GPU resource pool adopts a 6-machine 48-card Tesla P40 GPU server cluster mode to provide hardware resource service for algorithm training, support unified management of a mixed cluster formed by multiple kinds of hardware, use GPU cards for distributing resources in granularity, support hardware resource state monitoring management, support load balancing, smoothly expand capacity and achieve reasonable scheduling of hardware resources.
Specifically, the storage resource pool adopts 3 cluster storage servers, has high performance, and can realize data interaction of 4.5 ten thousand per second for 35kB pictures; the method supports the scalability, supports hundreds of clusters to the maximum, and linearly expands the concurrency capability; the method has high availability, supports double copying of metadata and user data, and ensures that the platform is stable and reliable.
The database layer designs corresponding database table structures and incidence relations thereof according to the training platform algorithm layer and the task layer, and provides data support for the training platform. The database records the incidence relation between a data set and a picture, data enhancement and calibration information, training project process information, training information of an algorithm model, model management and verification information, and intuitive page display and tracking can be performed on the power AI model from building to training to final model management and field deployment through the information.
The algorithm layer is shown in fig. 3, and the invention completes multi-algorithm unified container management and containerized heterogeneous resource unified scheduling.
Specifically, containerization management comprises a training frame container and a training tool container, and the training frame has mainstream deep learning frames such as: caffe, Tensorflow, Pyorch and the like, and the training tools are mainly divided into three categories of picture classification, target detection and image segmentation.
The task layer is divided into three categories of image classification, target detection and image segmentation according to an AI application scene of the power system, and algorithms under corresponding training tools can be selected according to different categories.
Specifically, the task creation, as shown in fig. 4, includes the following steps:
s1: logging in an electric power AI training platform through a Web account, and selecting a scene type (classification, detection and segmentation) to be trained;
s2, selecting a data set or uploading the data set to support online annotation;
s3, selecting corresponding training frame and AI algorithm model;
s4, uploading the algorithm configuration file to set the hyper-parameters, starting the training script to train, carrying out algorithm model verification, and modifying the hyper-parameters according to the verification information to carry out optimization;
and S5, uploading the model to a deployment application platform or a front-end device.
Specifically, the data management supports selection of multiple data enhancement modes and supports comparison and display of multi-column effects.
Specifically, the model training selects the number of GPU training according to GPU resource allocation, the training process is monitored, and the current training batch and the Loss curve are displayed on a page.
Specifically, model verification, the page shows the effect of the model on a verification set, multi-model simultaneous verification is supported, the model effect is contrastingly displayed, and result pictures are displayed in multiple columns.
Specifically, model management refers to a trigger mechanism which has version information of a model, a test report of the model and supports uploading and downloading of the model and iterative training.
The test report of the model comprises the overall evaluation of the model on the verification set, performance indexes such as a missing report rate, a false report rate, an mAP value and the like, and comprises an overall Loss curve and a correct rate curve of the model.
The model uploading and downloading refers to supporting the model to be downloaded locally and deployed to a field application platform.
The iterative training mechanism is that after field data is newly added to a set value, a model can be iteratively trained for a set number of times on the basis of an original model, the field detection model is guaranteed to be optimal, and the model is guaranteed to be rapidly deployed.
The synthesis of the above: the invention is based on containerization technology, realizes the customization of the electric power AI training platform by means of Web interface access, can meet the requirements of various service scenes, realizes the rapid training and application deployment of the algorithm in the electric power scene by using strong hardware resource allocation, can visually display the optimization process and performance indexes of the model, effectively manages the algorithm model, has an iterative training triggering mechanism, can realize rapid iteration of the model, and supports the dynamic update of the on-site algorithm.
While the invention has been described above with reference to an embodiment, various modifications may be made and equivalents may be substituted for elements thereof without departing from the scope of the invention. In particular, the various features of the embodiments disclosed herein may be used in any combination, provided that there is no structural conflict, and the combinations are not exhaustively described in this specification merely for the sake of brevity and conservation of resources. Therefore, it is intended that the invention not be limited to the particular embodiments disclosed, but that the invention will include all embodiments falling within the scope of the appended claims.
Claims (6)
1. The utility model provides an electric power AI training platform based on containerization management which characterized in that: the multi-language image recognition system comprises a hardware resource layer, an operating system layer, a communication layer, a database layer, an algorithm layer and a task layer, wherein the hardware resource layer comprises a GPU resource pool and a storage resource pool, the communication layer adopts RabbitMQ multi-language communication, the database layer designs a corresponding database table structure and an incidence relation thereof according to a training platform algorithm layer and the task layer, the algorithm layer presets a plurality of algorithm model image files based on deep learning frames such as Tensorflow, Pythrch, Caffe and the like, the task layer comprises three deep learning fields of image classification, target detection and image segmentation, the database layer has the functions of data set management and data calibration, has the functions of uploading and downloading, and supports data enhancement mode selection and multi-enhancement mode image contrast display.
2. The training method of the containerization management-based electric power AI training platform of claim 1, wherein: the training method of the electric power AI training platform based on containerization management comprises the following steps:
s1: creating a project, selecting a training data set, and selecting a built-in algorithm model;
s2: algorithm training, namely selecting a data set label to be trained, the number of GPUs, uploading an algorithm configuration file, and starting a training script;
s3: monitoring the state, namely monitoring the training process and displaying the current training batch and the Loss curve;
s4: after training is finished, storing the optimal model, and displaying a final Loss curve;
s5: and (4) performing algorithm verification, displaying the effect of the model on the verification set, supporting simultaneous verification of multiple models, comparing and displaying the model effect, and displaying the result pictures in multiple columns.
3. The training method of the containerization-management-based electric power AI training platform of claim 2, wherein: the method has the function of multi-model management, version information of the model, test reports of the model and a triggering mechanism for uploading and downloading of the model and iterative training.
4. The training method of the containerization-management-based electric power AI training platform of claim 3, wherein: the method comprises the overall evaluation of the model on a verification set, and performance indexes such as a missing report rate, a false report rate, an mAP value and the like, and comprises an overall Loss curve and a correct rate curve of the model.
5. The training method of the containerization-management-based electric power AI training platform of claim 4, wherein: and uploading and downloading of the model, and supporting downloading of the model to the local and deploying the model to a field application platform.
6. The training method of the containerization-management-based electric power AI training platform of claim 5, wherein: after the field data of the iterative training triggering mechanism is newly increased to a set value, the model can be iteratively trained for a set number of times on the basis of the original model, so that the optimization of a field detection model is ensured, and the rapid deployment of the model is ensured.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010823710.XA CN111814911A (en) | 2020-08-17 | 2020-08-17 | Electric power AI training platform based on containerization management and training method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010823710.XA CN111814911A (en) | 2020-08-17 | 2020-08-17 | Electric power AI training platform based on containerization management and training method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111814911A true CN111814911A (en) | 2020-10-23 |
Family
ID=72860275
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010823710.XA Pending CN111814911A (en) | 2020-08-17 | 2020-08-17 | Electric power AI training platform based on containerization management and training method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111814911A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112462779A (en) * | 2020-11-30 | 2021-03-09 | 汕头大学 | Group robot dynamic capture control method and system based on gene regulation network |
| CN112529078A (en) * | 2020-12-07 | 2021-03-19 | 杭州海康威视数字技术股份有限公司 | Service processing method, device and equipment |
| CN112926736A (en) * | 2021-02-26 | 2021-06-08 | 全球能源互联网研究院有限公司 | Deep learning technology tool chain system |
| CN113377387A (en) * | 2021-06-28 | 2021-09-10 | 中煤能源研究院有限责任公司 | Method for uniformly releasing, deploying and upgrading intelligent application of coal mine |
| CN113886026A (en) * | 2021-12-07 | 2022-01-04 | 中国电子科技集团公司第二十八研究所 | Intelligent modeling method and system based on dynamic parameter configuration and process supervision |
| CN114528106A (en) * | 2022-02-17 | 2022-05-24 | 西安电子科技大学 | Method for accelerating decision tree training by using GPU (graphics processing Unit) in radar signal sorting |
| CN117785408A (en) * | 2023-12-19 | 2024-03-29 | 中科城市大脑数字科技(无锡)有限公司 | AI model full life cycle management method, platform and storage medium |
-
2020
- 2020-08-17 CN CN202010823710.XA patent/CN111814911A/en active Pending
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112462779A (en) * | 2020-11-30 | 2021-03-09 | 汕头大学 | Group robot dynamic capture control method and system based on gene regulation network |
| CN112462779B (en) * | 2020-11-30 | 2023-07-25 | 汕头大学 | Dynamic group robot trapping control method and system based on gene regulation network |
| CN112529078A (en) * | 2020-12-07 | 2021-03-19 | 杭州海康威视数字技术股份有限公司 | Service processing method, device and equipment |
| CN112926736A (en) * | 2021-02-26 | 2021-06-08 | 全球能源互联网研究院有限公司 | Deep learning technology tool chain system |
| CN112926736B (en) * | 2021-02-26 | 2023-12-08 | 国网智能电网研究院有限公司 | Deep learning technology tool chain system |
| CN113377387A (en) * | 2021-06-28 | 2021-09-10 | 中煤能源研究院有限责任公司 | Method for uniformly releasing, deploying and upgrading intelligent application of coal mine |
| CN113377387B (en) * | 2021-06-28 | 2024-05-17 | 中煤能源研究院有限责任公司 | Unified publishing, deploying and upgrading method for intelligent coal mine application |
| CN113886026A (en) * | 2021-12-07 | 2022-01-04 | 中国电子科技集团公司第二十八研究所 | Intelligent modeling method and system based on dynamic parameter configuration and process supervision |
| CN113886026B (en) * | 2021-12-07 | 2022-03-15 | 中国电子科技集团公司第二十八研究所 | Intelligent modeling method and system based on dynamic parameter configuration and process supervision |
| CN114528106A (en) * | 2022-02-17 | 2022-05-24 | 西安电子科技大学 | Method for accelerating decision tree training by using GPU (graphics processing Unit) in radar signal sorting |
| CN114528106B (en) * | 2022-02-17 | 2024-05-17 | 西安电子科技大学 | Method for accelerating decision tree training by using GPU in radar signal sorting |
| CN117785408A (en) * | 2023-12-19 | 2024-03-29 | 中科城市大脑数字科技(无锡)有限公司 | AI model full life cycle management method, platform and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111814911A (en) | Electric power AI training platform based on containerization management and training method thereof | |
| US10963313B2 (en) | Automated reinforcement-learning-based application manager that learns and improves a reward function | |
| US11009836B2 (en) | Apparatus and method for optimizing quantifiable behavior in configurable devices and systems | |
| Yuan et al. | A data placement strategy in scientific cloud workflows | |
| CN112560244B (en) | Virtual simulation experiment system and method based on Docker | |
| CN105830049B (en) | Automation experiment platform | |
| Bird et al. | Update of the Computing Models of the WLCG and the LHC Experiments | |
| CN106020715A (en) | Storage pool capacity management | |
| US20140122387A1 (en) | Portable workload performance prediction for the cloud | |
| CN109219943A (en) | The automatic configuration of Machine To Machine system | |
| CN112433819A (en) | Heterogeneous cluster scheduling simulation method and device, computer equipment and storage medium | |
| CN110659741A (en) | AI model training system and method based on piece-splitting automatic learning | |
| CN104486255A (en) | Service resource dispatching method and device | |
| CN102722381A (en) | Techniques to optimize upgrade tasks | |
| CN102624865A (en) | Cluster load prediction method and distributed cluster management system | |
| WO2023020355A1 (en) | Distributed training method for ai model and related device | |
| CN109492070A (en) | A kind of city macroscopic view scene three-dimensional visualization plateform system | |
| US20200065703A1 (en) | Adversarial automated reinforcement-learning-based application-manager training | |
| US10552399B2 (en) | Predicting index fragmentation caused by database statements | |
| WO2015130643A1 (en) | Technologies for cloud data center analytics | |
| CN212541377U (en) | Electric power AI training platform based on containerization management | |
| US20200065701A1 (en) | Automated reinforcement-learning-based application manager that uses action tags and metric tags | |
| CN116126415A (en) | Scene-based cloud component configuration recommendation method and system | |
| CN114510319A (en) | Kubernetes cluster GPU space sharing method | |
| CN113296913A (en) | Data processing method, device and equipment based on single cluster and storage medium |
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 |