CN112380022A - Unmanned ship autonomous learning system, method and computer readable storage medium - Google Patents
Unmanned ship autonomous learning system, method and computer readable storage medium Download PDFInfo
- Publication number
- CN112380022A CN112380022A CN202011425161.7A CN202011425161A CN112380022A CN 112380022 A CN112380022 A CN 112380022A CN 202011425161 A CN202011425161 A CN 202011425161A CN 112380022 A CN112380022 A CN 112380022A
- Authority
- CN
- China
- Prior art keywords
- model
- loader
- task manager
- trained
- autonomous 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.)
- Pending
Links
Images
Classifications
-
- 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]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Stored Programmes (AREA)
Abstract
The invention provides an unmanned ship autonomous learning system, a method and a computer readable storage medium, wherein the system comprises a multitask request module, a Tensorflow Serving service module and a task management module, wherein the multitask request module is used for receiving access request information sent by a client and sending the request information to the Tensorflow Serving service module; the Tensorflow Serving service module is deployed in a Docker open source application container engine and is used for receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model and sending a calculation result to a client; and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end. The invention combines the Tensflow Serving service and a Docker open source application container engine, and deploys the Tensflow Serving service in a Docker container at a boat end. After the boat-end program sends a request, the calling and calculation of the model are both in a Docker container, and after the operation is finished, the result is returned to the boat-end receiving program, so that the system overhead is greatly reduced.
Description
Technical Field
The invention relates to the technical field of continuous autonomous learning of unmanned systems, in particular to an unmanned ship autonomous learning system, method and computer readable storage medium.
Background
The unmanned ship is a modern intelligent ship and is an unmanned intelligent platform sailing on the water surface in a semi-autonomous or fully-autonomous mode. In the process of executing tasks, the unmanned surface vehicle can be used for carrying out personnel material transferring and conveying, ocean exploration, maritime search and rescue, surveying and mapping tasks and the like at sea. Because the unmanned ship has diversified functions and has a requirement for autonomously executing tasks to a certain degree, various task-based intelligent algorithms should be deployed in an unmanned ship software system, and meanwhile, the intelligent algorithms are required to have growth performance, so that the autonomous capacity of the unmanned ship can be improved through updating and verification of the algorithms. The continuous autonomous learning technology of the shore-sea cooperation and water surface unmanned system is one of important ways for improving the autonomous capability of the unmanned ship, and with the development of emerging technologies such as machine learning, the autonomous capability of the machine can be improved through a large amount of training. Therefore, it is a trend of future development that unmanned boat software systems improve their autonomy by continuously learning and upgrading autonomously.
In order to deploy and integrate an intelligent algorithm on an unmanned ship to realize autonomous navigation of the unmanned ship, and simultaneously utilize a shore-sea cooperation and water surface unmanned system continuous autonomous learning system to perform algorithm model deployment, management and verification, the generally adopted intelligent algorithm deployment and integration method mainly comprises the steps of generating Dockers by the intelligent algorithms, deploying the Dockers on the unmanned ship, then mounting the Dockers on a model file directory, and realizing model calling, model version management and deployment by the algorithm. The drawbacks of this approach are:
the unmanned ship client has high calculation cost and higher requirements on equipment performance. The model deployment and calling process of the method is realized by the algorithm at the unmanned ship client, so that the environment of a program and a model needs to be established at the client in advance, and the method has higher requirements on the configuration and the performance of client hardware. Since the marine environment is worse than the land and has strict requirements on the volume and weight of the hull, the configuration and the operation performance of the hardware system of the unmanned ship are restricted, and the excessive calculation overhead affects the efficiency of program execution and even causes higher delay. In addition, model deployment and version management are difficult, and the model deployment and calling process of the method is realized by the algorithm at the unmanned ship client, so that the whole process is required to be completely automatic and stably executed, and an algorithm developer is required to perform additional optimization and adjustment on the model deployment and calling process.
Disclosure of Invention
The present invention has been made to overcome the above-mentioned problems, and the present invention provides an unmanned boat autonomous learning system, method, and computer-readable storage medium. The system combines the Tensflow Serving service with a Docker open source application container engine, and deploys the Tensflow Serving service in a Docker container at a boat end. After the boat-end program sends a request, the calling and calculation of the model are both in a Docker container, and after the operation is finished, the result is returned to the boat-end receiving program, so that the system overhead is greatly reduced. The problem of unmanned ship hardware system's configuration and operating performance receive the restriction among the prior art, too big calculation cost influences the efficiency of program execution, leads to higher delay even is solved.
The above object of the present invention can be achieved by the following technical solutions:
the invention provides an unmanned ship autonomous learning system, which comprises:
the multitask request module is used for receiving access request information sent by a client and sending the request information to the Tensorflow Serving service module;
the Tensorflow Serving service module is deployed in a Docker open source application container engine and is used for receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model and sending a calculation result to a client;
and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end.
Further, the tensrflow Serving service module includes:
the file monitoring component is used for monitoring a file system, searching and reading a multi-model file directory, creating a model loader for a model when the model of an updated version is found, and sending the model loader to the task manager after the creation of the loader is completed;
a model loader for pointing to an updated version of a model stored on a disk;
and the task manager is used for receiving the model of the updated version pointed in the model loader and then performing model service.
Further, when the task manager performs model service, whether a model is firstly pushed and deployed and whether resources required by the model are available are judged, and if the model is firstly deployed and corresponding resources are obtained, the task manager gives authority to the model loader to load the model;
or if the task manager judges that the model is online, when the online model is subjected to version updating, the task manager firstly queries the version management plug-in and then determines the updating mode of the loaded model according to the selected updating mode.
Further, the determining an update mode of the loading model according to the selected update mode includes:
if the priority maintenance availability is selected, the task manager enables the model loader to instantiate a new calculation graph and new weights, the online model and the model of the updated version are loaded at the moment, and the task manager unloads the online model after ensuring that the new version model can be safely serviced;
or, if the resources are selected to be maintained, the task manager does not apply for additional resources for the updated version of the model.
Further, the trained model comprises one of the following formats: pb format, ckpt format, and meta format.
Furthermore, Docker deployment is supported, the multi-model file directory is mounted under the virtual directory of Docker, and the Docker mirror image is pulled.
The invention also provides an unmanned boat autonomous learning method, which comprises the following steps:
receiving access request information sent by a client, and sending the request information to a Tensorflow Serving service module; wherein the Tensorflow Serving service module is deployed in a Docker open-source application container engine
Receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model, and sending a calculation result to a client; and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end.
Further, the receiving of the request information sent by the multitask request module, the selecting of the trained model from the multiple-model file directory, and the loading and calculating of the trained model include:
searching and reading a multi-model file directory, creating a model loader for a model when the model of an updated version is found, and sending the model loader to a task manager after creation of the loader is completed; wherein the model loader is to point to an updated version of a model stored on disk;
and after receiving the directed updated version model in the model loader, the task manager loads the model and performs corresponding model service according to the selected loading mode.
Further, the performing of the corresponding model service according to the selected loading manner includes the steps of:
judging whether the model is pushed and deployed for the first time and whether resources required by the model are available, and if the model is pushed and deployed for the first time and corresponding resources are obtained, giving authority to a model loader by a task manager to load the model;
or if the task manager judges that the model is online, when the online model is subjected to version updating, the task manager firstly queries the version management plug-in and then determines the updating mode of the loaded model according to the selected updating mode.
A third aspect of the invention provides a computer readable storage medium for storing a computer program which, when executed by a processor, causes the processor to perform the steps of any of the methods described above.
The invention has at least the following characteristics and advantages:
the invention provides a Tensorflow Serving-based continuous autonomous learning system for a shore-sea coordination unmanned surface vehicle, which comprises a multitask request module, a Tensorflow Serving service module and a data processing module, wherein the multitask request module is used for receiving access request information sent by a client and sending the request information to the Tensorflow Serving service module; the Tensorflow Serving service module is deployed in a Docker open source application container engine and is used for receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model and sending a calculation result to a client; and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end. The invention combines the Tensflow Serving service and a Docker open source application container engine, and deploys the Tensflow Serving service in a Docker container at a boat end. After the boat-end program sends a request, the calling and calculation of the model are both in a Docker container, and after the operation is finished, the result is returned to the boat-end receiving program, so that the system overhead is greatly reduced.
Further, in order to successfully load and deploy the model constructed by the Tensorflow, it is necessary to ensure that the format of the exported model is correct. The Tensorflow framework-based training comprises a plurality of model construction modes, and model files in pb format, ckpt format and meta format can be generated respectively. The invention adopts the SaveModel class provided by Tensorflow to generate the model, and the model can be stored in pb format. The SavedModel class allows one or more computation graphs to be stored simultaneously, which may allow us to save different computation graphs for different tasks. Meanwhile, the model file generated by the SavedModel supports multi-language loading, and the universality is stronger.
Furthermore, the method and the device can realize that the model can be loaded only by sending the request according to the requirement when the client calls the model, and manual loading is not needed.
Furthermore, the system aims at development requirements of intellectualization, autonomy and growth of the unmanned surface vehicle, and a continuous autonomous learning architecture of the unmanned surface vehicle based on bank-sea cooperation is patented, so that a data set training and verification platform of an unmanned surface vehicle intelligent algorithm is constructed, deployment integration and updating of relevant intelligent algorithms of the unmanned surface vehicle based on offshore tasks are realized, basic algorithm support is provided for intelligent unmanned surface products, and a technical foundation is laid for the intelligent level of the unmanned surface vehicle. The platform provides a data set for algorithm training by collecting data aiming at task characteristics and task target characteristics of the unmanned surface vehicle, and trains and updates various types of intelligent algorithms. The autonomous execution task of the unmanned ship on the sea is realized by utilizing the autonomous characteristics of the unmanned ship and combining an intelligent algorithm.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic block diagram of a structure of an unmanned ship autonomous learning system according to an embodiment of the present invention;
fig. 2 is a structural diagram of each component and interaction mode of a tensflo Serving service module of an unmanned ship autonomous learning system according to an embodiment of 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 embodiments described below 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.
Fig. 1 is a schematic block diagram of a structure of an unmanned ship autonomous learning system according to an embodiment of the present invention, including:
the multitask request module is used for receiving access request information sent by a client and sending the request information to the Tensorflow Serving service module;
the Tensorflow Serving service module is deployed in a Docker open source application container engine and is used for receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model and sending a calculation result to a client;
and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end.
In the specific embodiment of the invention, the model deployment supports Docker deployment, the model directory is mounted under the virtual directory of Docker, and the Docker mirror image is pulled.
The invention combines the continuous autonomous learning system architecture of the shore-sea protocol and water surface unmanned system and the application of Docker open source application container engine, carries out development and debugging based on a Tensorflow machine learning algorithm framework, generates a model by platform end training and pushes the model to a client, and deploys and stores the model in a client intelligent algorithm model by introducing Tensorflow service provided by Tensorflow authorities. The Tensflow Serving service and a Docker open source application container engine are combined, and the Tensflow Serving service is deployed in a Docker container at a boat end. After the boat-end program sends a request, the calling and calculation of the model are both in a Docker container, and after the operation is finished, the result is returned to the boat-end receiving program, so that the system overhead is greatly reduced.
Referring to fig. 2, a structural diagram of each component and interaction mode of a tensoflow Serving service module of an unmanned ship autonomous learning system according to an embodiment of the present invention is provided.
The tensrflow Serving service module specifically includes: the file monitoring component is used for monitoring a file system, searching and reading a multi-model file directory, creating a model loader for a model when the model of an updated version is found, and sending the model loader to the task manager after the creation of the loader is completed;
a model loader for pointing to an updated version of a model stored on a disk;
and the task manager is used for receiving the model of the updated version pointed in the model loader and then performing model service.
After the Tensorflow Serving service deployed on the continuous autonomous learning system reads the model file on disk, the cycle of the model service begins. The file monitoring component is used for monitoring the file system, searching and reading the model files at the same time, and creating a loader for the model after finding the new version of the model file each time. When the model loader is called, the relevant information of the model, such as the mode of loading the model, the requested memory, the GPU resource and the like, needs to be input. The model loader can point to a model that is connected to storage on disk, containing the relevant metadata needed to load the model. After the loader creation is complete, the file monitor component sends it to the task manager as the version to be loaded. And the task manager performs model service after receiving the version to be loaded.
When the task manager performs model service, judging whether a model is firstly pushed and deployed and whether resources required by the model are available, if so, giving authority to a model loader to load the model;
or if the task manager judges that the model is online, when the online model is subjected to version updating, the task manager firstly queries the version management plug-in and then determines the updating mode of the loaded model according to the selected updating mode.
When loading a new model, it may be preferable to maintain availability or resources. Maintaining availability means that the system tends to ensure that the system can always respond to client requests, and the task manager will cause the model loader to instantiate new computation graphs and new weights. At the moment, the two versions of the model are loaded, and the task manager unloads the original version model after ensuring that the new version model can be safely served. Keeping resources means that no additional resources are applied for the new version model, thereby achieving the effect of saving resources.
In order to successfully load and deploy the model built by the Tensorflow, the format of the exported model needs to be ensured to be correct. The Tensorflow framework-based training comprises a plurality of model construction modes, and model files in pb format, ckpt format and meta format can be generated respectively. The invention adopts the SaveModel class provided by Tensorflow to generate the model, and the model can be stored in pb format. The SavedModel class allows one or more computation graphs to be stored simultaneously, which may allow us to save different computation graphs for different tasks. Meanwhile, the model file generated by the SavedModel supports multi-language loading, and the universality is stronger.
The invention combines the Tensflow Serving service and a Docker open source application container engine, and deploys the Tensflow Serving service in a Docker container at a boat end. After the boat-end program sends a request, the calling and calculation of the model are both in a Docker container, and after the operation is finished, the result is returned to the boat-end receiving program, so that the system overhead is greatly reduced. Then, the invention can load the model only by sending the request according to the requirement when the client calls the model, and does not need manual loading. Finally, the invention automatically calls the model of the latest version of each algorithm without manual calling, or the algorithm itself filters and calls the latest version of the model. The invention supports multi-model deployment integration and meets the multi-task requirements of the unmanned surface vehicle.
The invention also provides an autonomous learning method of the unmanned ship system, which comprises the following steps:
receiving access request information sent by a client, and sending the request information to a Tensorflow Serving service module; wherein the Tensorflow Serving service module is deployed in a Docker open-source application container engine
Receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model, and sending a calculation result to a client; and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end.
Further, the receiving of the request information sent by the multitask request module, the selecting of the trained model from the multiple-model file directory, and the loading and calculating of the trained model include:
searching and reading a multi-model file directory, creating a model loader for a model when the model of an updated version is found, and sending the model loader to a task manager after creation of the loader is completed; wherein the model loader is to point to an updated version of a model stored on disk;
and after receiving the directed updated version model in the model loader, the task manager loads the model and performs corresponding model service according to the selected loading mode.
Further, the performing of the corresponding model service according to the selected loading manner includes the steps of:
judging whether the model is pushed and deployed for the first time and whether resources required by the model are available, and if the model is pushed and deployed for the first time and corresponding resources are obtained, giving authority to a model loader by a task manager to load the model;
or if the task manager judges that the model is online, when the online model is subjected to version updating, the task manager firstly queries the version management plug-in and then determines the updating mode of the loaded model according to the selected updating mode.
A third aspect of the invention provides a computer readable storage medium for storing a computer program which, when executed by a processor, causes the processor to perform the steps of any of the methods described above.
Although the present invention has been described with reference to the preferred embodiments, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. An unmanned boat autonomous learning system, comprising:
the multitask request module is used for receiving access request information sent by a client and sending the request information to the Tensorflow Serving service module;
the Tensorflow Serving service module is deployed in a Docker open source application container engine and is used for receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the model and sending a calculation result to a client;
and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end.
2. The unmanned boat autonomous learning system of claim 1, wherein the Tensorflow Serving service module comprises:
the file monitoring component is used for monitoring a file system, searching and reading a multi-model file directory, creating a model loader for a model when the model of an updated version is found, and sending the model loader to the task manager after the creation of the loader is completed;
a model loader for pointing to an updated version of a model stored on a disk;
and the task manager is used for receiving the model of the updated version pointed in the model loader and then performing model service.
3. The unmanned ship autonomous learning system of claim 2, wherein when the task manager performs model service, it determines whether a model is first deployed and whether resources required by the model are available, and if the model is first deployed and corresponding resources are acquired, the task manager gives permission to the model loader to perform model loading;
or if the task manager judges that the model is online, when the online model is subjected to version updating, the task manager firstly queries the version management plug-in and then determines the updating mode of the loaded model according to the selected updating mode.
4. The unmanned boat autonomous learning system of claim 3, wherein the determining an update mode of the loading model according to the selected update mode comprises:
if the priority maintenance availability is selected, the task manager enables the model loader to instantiate a new calculation graph and new weights, the online model and the model of the updated version are loaded at the moment, and the task manager unloads the online model after ensuring that the new version model can be safely serviced;
or, if the resources are selected to be maintained, the task manager does not apply for additional resources for the updated version of the model.
5. The unmanned boat autonomous learning system of claim 1, wherein the trained model comprises one of the following formats: pb format, ckpt format, and meta format.
6. The unmanned boat autonomous learning system of claim 1, wherein Docker deployment is supported, the multi-model file directory is mounted under a virtual directory of Docker, and a Docker mirror image is pulled.
7. An unmanned ship autonomous learning method is characterized by comprising the following steps:
receiving access request information sent by a client, and sending the request information to a Tensorflow Serving service module; wherein the Tensorflow Serving service module is deployed in a Docker open-source application container engine
Receiving request information sent by the multitask request module, selecting a trained model from a multi-model file directory, loading and calculating the trained model, and sending a calculation result to a client; and the multi-model file directory is used for mounting the intelligent algorithm model which is trained and completed by the platform end.
8. The unmanned boat autonomous learning method of claim 7, wherein the receiving of request information sent by the multitask request module, the selecting of trained models in a multi-model file directory, and the loading and computing of the trained models comprises the steps of:
searching and reading a multi-model file directory, creating a model loader for a model when the model of an updated version is found, and sending the model loader to a task manager after creation of the loader is completed; wherein the model loader is to point to an updated version of a model stored on disk;
and after receiving the directed updated version model in the model loader, the task manager loads the model and performs corresponding model service according to the selected loading mode.
9. The unmanned boat autonomous learning method of claim 7, wherein performing corresponding model services according to the selected loading mode comprises:
judging whether the model is pushed and deployed for the first time and whether resources required by the model are available, and if the model is pushed and deployed for the first time and corresponding resources are obtained, giving authority to a model loader by a task manager to load the model;
or if the task manager judges that the model is online, when the online model is subjected to version updating, the task manager firstly queries the version management plug-in and then determines the updating mode of the loaded model according to the selected updating mode.
10. A computer-readable storage medium for storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method of any one of claims 7 to 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011425161.7A CN112380022A (en) | 2020-12-09 | 2020-12-09 | Unmanned ship autonomous learning system, method and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011425161.7A CN112380022A (en) | 2020-12-09 | 2020-12-09 | Unmanned ship autonomous learning system, method and computer readable storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112380022A true CN112380022A (en) | 2021-02-19 |
Family
ID=74589404
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011425161.7A Pending CN112380022A (en) | 2020-12-09 | 2020-12-09 | Unmanned ship autonomous learning system, method and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112380022A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150160938A1 (en) * | 2013-12-10 | 2015-06-11 | Fmr Llc | Automation in rule engine in an integration development environment |
CN107329829A (en) * | 2017-06-28 | 2017-11-07 | 快云信息科技有限公司 | A kind of Service Source management-control method for application container engine of increasing income |
CN110516090A (en) * | 2019-08-09 | 2019-11-29 | 广东浪潮大数据研究有限公司 | A kind of object detecting method, device, equipment and computer readable storage medium |
CN112015519A (en) * | 2020-08-28 | 2020-12-01 | 江苏银承网络科技股份有限公司 | Model online deployment method and device |
-
2020
- 2020-12-09 CN CN202011425161.7A patent/CN112380022A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150160938A1 (en) * | 2013-12-10 | 2015-06-11 | Fmr Llc | Automation in rule engine in an integration development environment |
CN107329829A (en) * | 2017-06-28 | 2017-11-07 | 快云信息科技有限公司 | A kind of Service Source management-control method for application container engine of increasing income |
CN110516090A (en) * | 2019-08-09 | 2019-11-29 | 广东浪潮大数据研究有限公司 | A kind of object detecting method, device, equipment and computer readable storage medium |
CN112015519A (en) * | 2020-08-28 | 2020-12-01 | 江苏银承网络科技股份有限公司 | Model online deployment method and device |
Non-Patent Citations (1)
Title |
---|
JUSTIN HO: ""TensorFlow Serving + Docker + Tornado机器学习模型生产级快速部署https://zhuanlan.zhihu.comp52096200"", 《知乎》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110413391B (en) | Deep learning task service quality guarantee method and system based on container cluster | |
US11514304B2 (en) | Continuously provisioning large-scale machine learning models | |
EP3636390A1 (en) | Robot cluster scheduling system | |
CN112272234B (en) | Platform management system and method for realizing edge cloud cooperation intelligent service | |
Tang et al. | A container based edge offloading framework for autonomous driving | |
US20090019450A1 (en) | Apparatus, method, and computer program product for task management | |
CN103530182A (en) | Working scheduling method and device | |
CN114327858A (en) | Cloud edge end distributed computing power cooperation method and system based on control domain | |
CN110990154B (en) | Big data application optimization method, device and storage medium | |
CN111459621B (en) | Cloud simulation integration and scheduling method and device, computer equipment and storage medium | |
CN115421735A (en) | Heterogeneous deployment method and device for deep learning task and electronic equipment | |
CN111625317A (en) | Container cloud construction method and related device of business system | |
CN112380022A (en) | Unmanned ship autonomous learning system, method and computer readable storage medium | |
CN108770014B (en) | Calculation evaluation method, system and device of network server and readable storage medium | |
KR102045997B1 (en) | Method for scheduling task in big data analysis platform based on distributed file system, program and computer readable storage medium therefor | |
CN114237830B (en) | Dispatching and physical machine capacity expansion method and system for quick start of virtual machine in target range | |
CN116149701A (en) | Online software upgrading method and system for edge terminal | |
CN114906160A (en) | Vehicle control and task processing method and device, computing equipment and system | |
CN112416591A (en) | Distributed task processing method, device, equipment, storage medium and system | |
CN116882523A (en) | Multi-task parallel joint learning engine architecture system | |
US20240070277A1 (en) | Cloud-based updating of root file systems using system partitioning | |
CN117519994B (en) | NUMA perception scheduling method and system based on k8s | |
CN116980423B (en) | Model scheduling method, device, computing system, equipment and readable storage medium | |
CN117056057A (en) | Algorithm scheduling method, device, electronic equipment and readable storage medium | |
WO2022256980A1 (en) | Source archive optimizations for reducing container image sizes |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210219 |
|
RJ01 | Rejection of invention patent application after publication |