WO2020082611A1

WO2020082611A1 - Method for carrying out deep learning on basis of blockchain platform and electronic device

Info

Publication number: WO2020082611A1
Application number: PCT/CN2019/070289
Authority: WO
Inventors: 金龙
Original assignee: 平安科技（深圳）有限公司
Priority date: 2018-10-25
Filing date: 2019-01-03
Publication date: 2020-04-30
Also published as: CN109409738A

Abstract

Disclosed in the present application are a method for carrying out deep learning on the basis of a blockchain platform, and an electronic device, the method comprising: receiving a deep learning model selected by a task distributor at a blockchain platform; receiving a training set inputted by the task distributor; generating a deep learning task according to the training set and the deep learning model; distributing the deep learning task to computing nodes in the blockchain platform; and carrying out deep learning for the deep learning task by means of the computing nodes to which the task is distributed. Employing the solution above solves the existing technical problems of the cost of deep learning being too high and the efficiency of training being low, and achieves the technical effects of reducing the cost of training and increasing the efficiency of training.

Description

Deep learning method and electronic device based on blockchain platform

Cross-reference of related applications

This application declares to enjoy the priority of the Chinese patent application submitted on October 25, 2018, with the application number CN 201811252535.2 and titled "Deep learning method and electronic device based on blockchain platform". Incorporated by reference in this application.

Technical field

The present application relates to the field of blockchain technology, in particular to a method for deep learning based on a blockchain platform, an electronic device, and a computer-readable storage medium.

Background technique

Deep learning requires the training of models through neural network calculations. However, the data model training process consumes a lot of computing resources. If artificial intelligence products want to achieve better product indicators, in addition to algorithms, massive amounts of data are required for training. However, the more data, in the case of equal computing resources, means longer training, for example: more than a week or even a month to several months.

If the parameters in the training process are incorrect, the training needs to be repeated. The training time is too long, which is extremely unfavorable for the iterative update of enterprise products. It is likely to fail in the industry competition, resulting in many manufacturers having to invest a lot of money to purchase GPU The cost of hardware resources such as FPGA is too high.

In response to the above problems, no effective solution has been proposed yet.

Summary of the invention

The purpose of the present application is to provide a method for deep learning based on a blockchain platform, an electronic device and a computer-readable storage medium, so as to solve the existing technical problem of excessively high cost of deep learning.

This application solves the above technical problems through the following technical solutions:

According to an aspect of this application, a method for deep learning based on a blockchain platform is provided, including the following steps:

Receive the deep learning model selected by the task publisher on the blockchain platform;

Receiving the training set input by the task publisher;

Generate a deep learning task according to the training set and the deep learning model;

Distribute the deep learning task to the computing nodes in the blockchain platform;

Perform deep learning on the deep learning task through the distributed computing nodes.

In one embodiment, distributing the deep learning task to the computing nodes in the blockchain platform includes:

Publish the deep learning task on the blockchain platform, wherein the release information carries the price corresponding to the unit task volume;

Receiving a claim request for a deep learning task issued by a computing node, where the claim request carries the amount of claimed tasks;

The computing node corresponding to the claim request is assigned a deep learning task of the requested task amount.

In one embodiment, publishing the deep learning task on the blockchain platform includes:

Divide the deep learning task into multiple unit task volumes;

Calculate the price of unit task volume;

Determine the computing nodes that can apply for the deep learning task according to the computing environment required by the deep learning task;

The deep learning task is issued to a computing node that can apply for the deep learning task.

In one embodiment, performing deep learning on the deep learning task through the distributed computing node includes:

The computing node receives a deep learning task distributed to itself, where the deep learning task distributed to itself carries: a deep learning model and a training set;

The computing node inputs the training set as training samples into the deep learning model for training to obtain the trained deep learning model;

Wherein, the computing nodes in the blockchain platform include at least one of the following: full-featured nodes running on the FPGA server cluster, idle GPU computing nodes in enterprises, and idle GPU computing nodes in individuals.

In one embodiment, after performing deep learning on the deep learning task through the distributed computing node, the method further includes:

Acquiring execution information and completion status of the deep learning task;

Record the execution information and the completion status in the form of transactions in the blockchain platform.

On the other hand, there is provided an electronic device including a memory and a processor for storing a deep learning system based on a blockchain platform that can be executed by the processor. Deep learning systems include:

The first receiving module is used to receive the deep learning model selected by the task publisher on the blockchain platform;

A second receiving module, configured to receive the training set input by the task publisher;

A generating module, configured to generate a deep learning task according to the training set and the deep learning model;

A distribution module for distributing the deep learning tasks to the computing nodes in the blockchain platform;

A deep learning module is used to perform deep learning on the deep learning task through the computing node to which it is distributed.

In one embodiment, the distribution module includes:

A release unit, configured to release the deep learning task on the blockchain platform, wherein the release information carries a price corresponding to a unit task volume;

A receiving unit, configured to receive a computing node's claim request for the issued deep learning task, where the claim request carries the claimed task amount;

The allocation unit is used for allocating the deep learning task of the requested task amount to the computing node corresponding to the claim request.

In one embodiment, the deep learning module includes:

The receiving unit is used for the computing node to receive the deep learning task distributed to itself, wherein the deep learning task distributed to itself carries: deep learning model and training set;

The input unit is used by the computing node to input the training set as a training sample into the deep learning model for training to obtain the trained deep learning model;

According to yet another aspect of the present application, a computer device is provided, including: a memory and a processor, wherein the processor is used to perform the steps of the above method.

According to yet another aspect of the present application, there is provided a non-volatile computer-readable storage medium on which computer instructions are stored, and when the instructions are executed, the steps of the above method are implemented.

The positive progress of this application lies in: receiving the deep learning model and input training set selected by the task publisher on the blockchain platform to generate a deep learning task; after the task publisher confirms the release of the deep learning task, the The deep learning task is distributed to the computing nodes in the blockchain platform for deep learning. In the above scheme, the training tasks are distributed and processed through the blockchain platform, which solves the existing technical problems of too high deep learning costs and low training efficiency, and achieves the technical effects of reducing training costs and improving training efficiency.

BRIEF DESCRIPTION

1 is a flowchart of a deep learning method based on blockchain according to an embodiment of the present application;

2 is a structural block diagram of a deep learning system based on blockchain according to an embodiment of the present application;

3 is a structural block diagram of a deep learning system based on blockchain according to an embodiment of the present application;

4 is a schematic diagram of an optional hardware architecture of an electronic device according to an embodiment of the present application.

detailed description

In order to make the purpose, technical solutions and advantages of the present application more clear, the following describes the present application in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. Based on the embodiments in this application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of this application.

Considering that the existing artificial products still need to be decoded by neural network calculation after going online, the greater the number of users, the greater the amount of calculation required, the higher the cost, and the frequency of user access in different time periods. As a result, purchasing large amounts of computing resources at one time will inevitably result in idle waste. The three elements of artificial intelligence are computing power, algorithms, and data. Among them, the amount of data is a very important factor that affects the indicators of artificial intelligence products. Companies that make artificial intelligence products need to continue to manually label low-quality data or directly purchase high-quality Data, but a lot of data not only involves user privacy, but also for the data provider also wants the data to be non-reproducible. It only sells the right to use the data and does not sell the ownership. However, if the data receiver does not have access to the data, then It is impossible to train based on data.

Based on this, a deep learning method and electronic device based on blockchain are provided in this example, and the method and the electronic device are specifically described as follows:

Example 1

The method for deep learning based on blockchain provided by the present application will be described below with reference to the drawings.

FIG. 1 is a schematic diagram of an optional process of the deep learning method based on blockchain in this application. As shown in FIG. 1, the method may include the following steps:

Step 101: Receive the deep learning model selected by the task publisher on the blockchain platform;

Specifically, before receiving the deep learning model selected by the task publisher on the blockchain platform, the deep learning model packaged and released by the user may be received, wherein the deep learning model carries: operating environment, input data format, output Data format; publish the deep learning model to the blockchain platform for the task publisher to choose.

In actual implementation, after researching and developing a deep learning model for an artificial intelligence application, researchers can open source and package and publish to the deep learning blockchain to provide the operating environment and input and output data format standards, and can choose to use public The data set submits training / test tasks. Among them, when the released deep learning model is used by others, the publisher can get corresponding rewards.

Step 102: Receive the training set input by the task publisher;

Step 103: Generate a deep learning task according to the training set and the deep learning model;

In order to enable the task publisher to know the cost and processing status of the training task, after generating the deep learning task according to the training set and the deep learning model, the task processing information of the deep learning task can be generated and displayed. The task processing information includes at least one of the following: estimated cost, available computing nodes, and unit cost; wherein, by dynamically adjusting the unit cost of all deep learning models, all computing nodes in the blockchain platform perform any The overall rewards for learning tasks are the same.

Step 104: Distribute the deep learning task to the computing nodes in the blockchain platform;

The computing nodes in the above-mentioned blockchain platform may include, but are not limited to, at least one of the following: full-featured nodes running on FPGA server clusters, idle GPU computing nodes in enterprises, and idle GPU computing nodes in individuals.

Step 105: Perform deep learning on the deep learning task through the distributed computing node.

Specifically, after performing deep learning on the deep learning task through the distributed computing node, the execution information and completion status of the deep learning task can be obtained; the execution information and the completion status are recorded in the form of a transaction In the blockchain platform, because it is recorded in the blockchain platform in the form of transactions, therefore, the processing status of these tasks can be traced and queried later.

In the actual implementation process, distributing the deep learning task to the computing nodes in the blockchain platform may include:

S1: The deep learning task is released on the blockchain platform, wherein the release information carries the price corresponding to the unit task volume;

S2: Receive a claim request for a deep learning task issued by a computing node, where the claim request carries the amount of claimed tasks;

S3: Assign the deep learning task of the requested task amount to the computing node corresponding to the claim request.

That is, the task is distributed to the claim platform, and the computing node makes the claim independently according to its own situation.

Specifically, during the release process, the deep learning task can be divided into multiple unit task volumes; the price of the unit task volume is calculated; and the deep learning task can be claimed according to the computing environment required by the deep learning task The computing node of publishes the deep learning task to the computing node that can claim the deep learning task.

For the computing node, performing deep learning on the deep learning task through the computing node distributed to may include:

S1: The computing node receives the deep learning task distributed to itself, where the deep learning task distributed to itself carries: a deep learning model and a training set;

S2: The computing node inputs the training set as a training sample into the deep learning model for training to obtain the trained deep learning model.

The above method will be described below in conjunction with a specific embodiment. However, it is worth noting that this specific embodiment is only for better description of the present application, and does not constitute an undue limitation on the present application.

In this example, a decentralized, low-cost, private deep learning computing platform based on blockchain technology is provided. The deep learning computing nodes on the chain can be composed of many forms, for example, it can include: large Full-featured nodes (permanent nodes) running on GPU or FPGA server clusters, idle GPU server computing nodes for small and medium-sized enterprises, and personal idle GPU computing nodes.

Specifically, the deep learning blockchain conducts transactions based on smart contracts, and incentivizes mining nodes according to the reward system designed by smart contracts. While ensuring the safe and stable operation of the system, it also allows all participants to get rewards from it, so that artificial intelligence manufacturers can obtain neural network deep learning computing power at a low cost.

Specifically, the deep learning blockchain can be designed according to the following principles:

1) Extension principle: each module is loosely coupled, it is easy to add new modules, and the update of each module itself does not need to depend on the changes of other module interfaces.

2) Scaling principle: Customer product user access is fluctuating. If a large number of users access a node, it will inevitably bring down the service of the node. Therefore, the container that sets the node itself can be automatically deployed. , You can quickly achieve horizontal expansion.

3) Privacy principle: All participants in the ecology, mining nodes, artificial intelligence manufacturers, data providers, etc. can be protected by privacy, and participants can selectively open according to their own needs.

The above-mentioned deep learning blockchain platform may include the following modules:

1) Deep learning calculation module:

The deep learning calculation module is used to complete: distribution of calculation tasks and distribution of container calculations. Specifically, after the computing container is successfully deployed, the verification calculation is performed. After the verification calculation is passed, the calculation task is distributed across the entire network through the blockchain network.

2) Calculation container:

It is used to manage the entire life cycle of container instances for a single user or use a group, provide virtual services according to user needs, and is responsible for container creation, suspension, suspension, adjustment, migration, restart, destruction and other operations.

When the computing capacity of a user's concurrent request reaches a specific value of the container allocation (this value can be set by the user), the container computing engine will start an alarm and will automatically deploy containers to other normal nodes for expansion.

Specifically, you can first read the configuration file, read the configuration parameters, and complete the initialization of the message queue according to the configuration for subsequent internal message interaction with other components. At the same time, the DB server is started according to the configuration items in the configuration file, and each API in the configuration file corresponds to a server. In addition, according to the number of GPU cores of the system n, each DB server will have n coroutines to process the request.

3) Image management system:

A set of virtual container image search and retrieval system, with functions of creating image, uploading image, deleting image, editing basic information of image. Image management system mainly consists of two services: image management API and image management register.

Among them, the mirror management API is the entrance of the mirror management system service, and is responsible for receiving user API requests. The mirror management register processes mirror metadata related requests. When the mirror management API receives the user's API request, if it is determined that the request is related to metadata, the request is transferred to the mirror management register service. Then, the mirror management register will parse the content of the user's metadata request, and interact with the database to access and update the metadata of the mirror.

Based on the above deep learning blockchain platform, deep learning can be performed as follows:

S1, after researchers have developed a deep learning model for an artificial intelligence application, open source and package it and release it to the deep learning blockchain, provide the operating environment and input and output data format standards, and can choose to submit training / test tasks with public data sets . Among them, when the released deep learning model is used by others, the publisher can get corresponding rewards.

S2. Training task release: The publisher selects the published model algorithm, and releases the training / test task after packing the data. Before submission, you can display: estimated cost consumption, available nodes, unit cost and other information. By dynamically adjusting the unit cost of all model algorithms, you can achieve the same overall return for all nodes in the entire network to perform any deep calculation task. Among them, the task execution status and task completion status are recorded in the block in the form of transactions.

S3, task allocation and collection: receive the tasks broadcast to the deep learning blockchain and the running status of other nodes, create a block every certain time and select the tasks to run according to the algorithm and execute them. In addition to the fees paid by the task publisher, the final reward of the node will be calculated in the blockchain after the task is completed, and distributed in proportion to the execution of each node.

In the above example, in order to solve the problems of huge resource consumption, inability to share, a certain degree of waste, and insufficient flexibility of deep learning clusters in existing deep learning clusters, a blockchain-based deep learning cluster system was proposed. In the chain network, platforms with deep learning resources and capabilities of all parties can be connected to the entire deep learning network, and rely on intelligent algorithms to arrange and schedule deep learning tasks, thereby effectively improving platform resource utilization. In addition, tasks compete for node capabilities, making the cluster more efficient. Because the blockchain system is established, the task is made private and information security is effectively protected. Further, according to the task scheduling fees generated by the blockchain platform, it is accounted and recognized by multiple parties, which improves trust. All the tasks in the network are coordinated by each deep learning node, which avoids the problem that the centralized node leaks the task content and results.

In the above embodiment, the deep learning model selected by the task publisher on the blockchain platform and the input training set are received to generate a deep learning task; after the task publisher confirms the release of the deep learning task, the depth The learning task is distributed to the computing nodes in the blockchain platform for deep learning. In the above scheme, the training tasks are distributed and processed through the blockchain platform, which solves the existing technical problems of too high deep learning costs and low training efficiency, and achieves the technical effects of reducing training costs and improving training efficiency.

Through the description of the above embodiments, those skilled in the art can clearly understand that the methods in the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware, but in many cases the former is better Implementation.

Example 2

Based on the deep learning method based on the blockchain platform provided in the first embodiment above, this embodiment provides a system for deep learning based on the blockchain platform. Specifically, FIG. 2 and FIG. 3 show the An optional structural block diagram of a deep learning system for a blockchain platform. The deep learning system based on a blockchain platform is divided into one or more program modules, and one or more program modules are stored in a storage medium. It is executed by one or more processors to complete this application. The program module referred to in this application refers to a series of computer program instruction segments that can perform specific functions. It is more suitable than the program itself to describe the execution process of a deep learning system based on a blockchain platform in a storage medium. The following description will specifically introduce Functions of each program module in this embodiment:

As shown in FIG. 2, a system 20 for deep learning based on a blockchain platform includes:

The first receiving module 201 is used to receive the deep learning model selected by the task publisher on the blockchain platform;

The second receiving module 202 is configured to receive the training set input by the task publisher;

A generating module 203, configured to generate a deep learning task according to the training set and the deep learning model;

The distribution module 204 is used to distribute the deep learning task to the computing nodes in the blockchain platform;

The deep learning module 205 is configured to perform deep learning on the deep learning task through the distributed computing node.

In one embodiment, the computing nodes in the above-mentioned blockchain platform may include, but are not limited to, at least one of the following: a full-featured node running on an FPGA server cluster, an idle GPU computing node in an enterprise, and an individual idle GPU computing node.

In one embodiment, as shown in FIG. 3, the above-mentioned system may further include: a third receiving module 301, configured to receive the depth of the packaged release by the user before receiving the deep learning model selected by the task publisher on the blockchain platform Learning model, wherein the deep learning model carries: operating environment, input data format, output data format; publishing module 302, used to publish the deep learning model to the blockchain platform, for the task publisher select.

In one embodiment, after generating a deep learning task according to the training set and the deep learning model, the above system may also generate task processing information of the deep learning task, where the task processing information includes at least one of the following One: estimated costs, available computing nodes, and unit costs; where, by dynamically adjusting the unit costs of all deep learning models, the overall return of all computing nodes in the blockchain platform to perform any learning task is the same ; Display the task processing information.

In one embodiment, after performing deep learning on the deep learning task through the computing node to which it is distributed, the above-mentioned system can also obtain execution information and completion status of the deep learning task; Recorded in the blockchain platform in the form of transactions.

In one embodiment, the above-mentioned distribution module may include: a publishing unit for publishing the deep learning task on the blockchain platform, wherein the publishing information carries a price corresponding to a unit task amount; a receiving unit is used for Receiving a claim request for a deep learning task issued by a computing node, wherein the claim request carries the claimed task amount; an allocation unit is configured to allocate a deep learning task requesting the claimed task amount to the computing node corresponding to the claim request.

In one embodiment, the publishing unit includes: a division subunit for dividing the deep learning task into a plurality of unit task amounts; a calculation subunit for calculating the price of the unit task amount; determining a subunit, using To determine the computing nodes that can apply for the deep learning task according to the computing environment required by the deep learning task; issue a subunit for publishing the deep learning task to the computing nodes that can apply for the deep learning task .

In one embodiment, the deep learning module is specifically used for the computing node to receive a deep learning task distributed to itself, where the deep learning task distributed to itself carries: a deep learning model and a training set; the computing node will The training set is input as a training sample into the deep learning model for training to obtain the trained deep learning model.

Regarding the device in the above embodiment, the specific manner in which the operations of each unit and module have been described in detail in the embodiment of the method, and will not be elaborated here.

In various implementations of this embodiment, the deep learning model selected by the task publisher on the blockchain platform and the input training set are received to generate a deep learning task; after the task publisher confirms to release the deep learning task, Distribute the deep learning task to the computing nodes in the blockchain platform for deep learning. In the above scheme, the training tasks are distributed and processed through the blockchain platform, which solves the existing technical problems of too high deep learning costs and low training efficiency, and achieves the technical effects of reducing training costs and improving training efficiency.

Example 3

An electronic device is provided in the third preferred embodiment of the present application. 4 is a schematic diagram of a hardware architecture according to an embodiment of an electronic device of the present application. In this embodiment, the electronic device 2 is a device that can automatically perform numerical calculation and / or information processing according to an instruction set or stored in advance. For example, it may be a smart phone, tablet computer, notebook computer, desktop computer, rack server, blade server, tower server, or rack server (including an independent server or a server cluster composed of multiple servers). As shown in the figure, the electronic device 2 includes at least but not limited to: a memory 21, a processor 22, a network interface 23, and a system 20 for deep learning based on a blockchain platform that can communicate with each other via a system bus. among them:

The memory 21 includes at least one type of computer-readable storage medium. The readable storage medium includes a flash memory, a hard disk, a multimedia card, a card-type memory (for example, SD or DX memory, etc.), a random access memory (RAM), and a static random access memory. (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 21 may be an internal storage module of the electronic device 2, such as a hard disk or a memory of the electronic device 2. In other embodiments, the memory 21 may also be an external storage device of the electronic device 2, such as a plug-in hard disk equipped on the electronic device 2, a smart memory card (Smart Media, Card, SMC), and secure digital (Secure Digital, SD) card, flash card (Flash Card), etc. Of course, the memory 21 may also include both the internal storage module of the electronic device 2 and its external storage device. In this embodiment, the memory 21 is generally used to store an operating system installed in the electronic device 2 and various application software, such as program codes of a system 20 for deep learning based on a blockchain platform. In addition, the memory 21 may also be used to temporarily store various types of data that have been output or will be output.

The processor 22 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip in some embodiments. The processor 22 is generally used to control the overall operation of the electronic device 2, such as performing control and processing related to data interaction or communication with the electronic device 2. In this embodiment, the processor 22 is used to run the program code stored in the memory 21 or process data, such as a system 20 for deep learning based on a blockchain platform.

The network interface 23 may include a wireless network interface or a wired network interface, and the network interface 23 is generally used to establish a communication connection between the electronic device 2 and other electronic devices. For example, the network interface 23 is used to connect the electronic device 2 with an external terminal through a network, and establish a data transmission channel and a communication connection between the electronic device 2 and the external terminal. The network can be Intranet, Internet, Global System of Mobile (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Wireless or wired networks such as Bluetooth and Wi-Fi.

It should be noted that FIG. 4 only shows an electronic device having components 21-23, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

In this embodiment, the system 20 for deep learning based on the blockchain platform stored in the memory 21 can also be divided into one or more program modules, one or more program modules are stored in the memory 21, and One or more processors (the processor 22 in this embodiment) are executed to complete the application.

Example 4

This embodiment also provides a computer-readable storage medium in which a system for deep learning based on a blockchain platform is stored. The system for monitoring and dialing tasks can be executed by at least one processor, so that at least one The processor executes the steps of the deep learning method based on the blockchain platform as in the first embodiment.

In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory ( ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium may be an internal storage unit of the computer device, such as a hard disk or memory of the computer device. In other embodiments, the computer-readable storage medium may also be an external storage device of the computer device, for example, a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media (SMC), Secure Digital (Secure Digital) , SD) card, flash card (Flash Card), etc. Of course, the computer-readable storage medium may also include both the internal storage unit of the computer device and the external storage device thereof. In this embodiment, the computer-readable storage medium is generally used to store the operating system and various application software installed on the computer device, for example, the program code of the customer guarantee analysis system in the second embodiment. In addition, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or are to be output.

Obviously, those skilled in the art should understand that the modules or steps in the embodiments of the present application described above can be implemented by a general-purpose computing device, and they can be concentrated on a single computing device or distributed among multiple computing devices. On the Internet, optionally, they can be implemented with program code executable by the computing device, so that they can be stored in the storage device and executed by the computing device, and in some cases, can be different from here The steps shown or described are executed in the order of, or they are made into individual integrated circuit modules separately, or multiple modules or steps among them are made into a single integrated circuit module for implementation. In this way, the embodiments of the present application are not limited to any specific combination of hardware and software.

The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the description and drawings of this application, or directly or indirectly used in other related technical fields , The same reason is included in the scope of patent protection of this application.

Claims

A method for deep learning based on a blockchain platform is characterized by the following steps:

Receive the deep learning model selected by the task publisher on the blockchain platform;

Receiving the training set input by the task publisher;

Generate a deep learning task according to the training set and the deep learning model;

Distribute the deep learning task to the computing nodes in the blockchain platform;

Perform deep learning on the deep learning task through the distributed computing nodes.
The method according to claim 1, wherein distributing the deep learning task to the computing nodes in the blockchain platform includes:

Publish the deep learning task on the blockchain platform, wherein the release information carries the price corresponding to the unit task volume;

Receiving a claim request for a deep learning task issued by a computing node, where the claim request carries the amount of claimed tasks;

The computing node corresponding to the claim request is assigned a deep learning task of the requested task amount.
The method of claim 2, wherein publishing the deep learning task on the blockchain platform includes:

Divide the deep learning task into multiple unit task volumes;

Calculate the price of unit task volume;

Determine the computing nodes that can apply for the deep learning task according to the computing environment required by the deep learning task;

The deep learning task is issued to a computing node that can apply for the deep learning task.
The method according to claim 1, wherein performing deep learning on the deep learning task through the distributed computing node includes:

The computing node receives a deep learning task distributed to itself, where the deep learning task distributed to itself carries: a deep learning model and a training set;

The computing node inputs the training set as training samples into the deep learning model for training to obtain the trained deep learning model;

Wherein, the computing nodes in the blockchain platform include at least one of the following: full-featured nodes running on the FPGA server cluster, idle GPU computing nodes in enterprises, and idle GPU computing nodes in individuals.
The method according to claim 1, wherein after performing deep learning on the deep learning task through the distributed computing node, the method further comprises:

Acquiring execution information and completion status of the deep learning task;

Record the execution information and the completion status in the form of transactions in the blockchain platform.
An electronic device includes a memory and a processor, wherein the memory is used to store a deep learning system based on a blockchain platform executable by the processor, and the deep learning based on the blockchain platform The system includes:

The first receiving module is used to receive the deep learning model selected by the task publisher on the blockchain platform;

A second receiving module, configured to receive the training set input by the task publisher;

A generating module, configured to generate a deep learning task according to the training set and the deep learning model;

A distribution module for distributing the deep learning tasks to the computing nodes in the blockchain platform;

A deep learning module is used to perform deep learning on the deep learning task through the computing node to which it is distributed.
The electronic device according to claim 6, wherein the distribution module comprises:

A release unit, configured to release the deep learning task on the blockchain platform, wherein the release information carries a price corresponding to a unit task volume;

A receiving unit, configured to receive a computing node's claim request for the issued deep learning task, where the claim request carries the claimed task amount;

The allocation unit is used for allocating the deep learning task of the requested task amount to the computing node corresponding to the claim request.
The electronic device according to claim 7, wherein publishing the deep learning task on the blockchain platform includes:

Divide the deep learning task into multiple unit task volumes;

Calculate the price of unit task volume;

Determine the computing nodes that can apply for the deep learning task according to the computing environment required by the deep learning task;

The deep learning task is issued to a computing node that can apply for the deep learning task.
The electronic device according to claim 6, wherein the deep learning module comprises:

The receiving unit is used for the computing node to receive the deep learning task distributed to itself, wherein the deep learning task distributed to itself carries: deep learning model and training set;

The input unit is used by the computing node to input the training set as a training sample into the deep learning model for training to obtain the trained deep learning model;

Wherein, the computing nodes in the blockchain platform include at least one of the following: full-featured nodes running on the FPGA server cluster, idle GPU computing nodes in enterprises, and idle GPU computing nodes in individuals.
The electronic device according to claim 6, characterized in that it is further used to obtain execution information and completion status of the deep learning task after performing deep learning on the deep learning task through the computing node to which it is distributed; The execution information and the completion status are recorded in the blockchain platform in the form of transactions.
A computer device includes: a memory and a processor, wherein the processor is used to perform the following steps:

Receive the deep learning model selected by the task publisher on the blockchain platform;

Receiving the training set input by the task publisher;

Generate a deep learning task according to the training set and the deep learning model;

Distribute the deep learning task to the computing nodes in the blockchain platform;

Perform deep learning on the deep learning task through the distributed computing nodes.
The computer device according to claim 11, wherein the distribution of the deep learning task to the computing nodes in the blockchain platform includes:

Publish the deep learning task on the blockchain platform, wherein the release information carries the price corresponding to the unit task volume;

Receiving a claim request for a deep learning task issued by a computing node, where the claim request carries the amount of claimed tasks;

The computing node corresponding to the claim request is assigned a deep learning task of the requested task amount.
The computer device according to claim 12, wherein publishing the deep learning task on the blockchain platform includes:

Divide the deep learning task into multiple unit task volumes;

Calculate the price of unit task volume;

Determine the computing nodes that can apply for the deep learning task according to the computing environment required by the deep learning task;

The deep learning task is issued to a computing node that can apply for the deep learning task.
The computer device according to claim 11, wherein performing deep learning on the deep learning task through the distributed computing node includes:

The computing node receives a deep learning task distributed to itself, where the deep learning task distributed to itself carries: a deep learning model and a training set;

The computing node inputs the training set as training samples into the deep learning model for training to obtain the trained deep learning model;

Wherein, the computing nodes in the blockchain platform include at least one of the following: full-featured nodes running on the FPGA server cluster, idle GPU computing nodes in enterprises, and idle GPU computing nodes in individuals.
The computer device according to claim 11, wherein after performing deep learning on the deep learning task through the distributed computing node, the method further comprises:

Acquiring execution information and completion status of the deep learning task;

Record the execution information and the completion status in the form of transactions in the blockchain platform.
A non-volatile computer-readable storage medium having computer instructions stored thereon, the instructions implement the following steps when executed:

Receive the deep learning model selected by the task publisher on the blockchain platform;

Receiving the training set input by the task publisher;

Generate a deep learning task according to the training set and the deep learning model;

Distribute the deep learning task to the computing nodes in the blockchain platform;

Perform deep learning on the deep learning task through the distributed computing nodes.
The non-volatile computer-readable storage medium of claim 16, wherein distributing the deep learning task to computing nodes in the blockchain platform includes:

Publish the deep learning task on the blockchain platform, wherein the release information carries the price corresponding to the unit task volume;

Receiving a claim request for a deep learning task issued by a computing node, where the claim request carries the amount of claimed tasks;

The computing node corresponding to the claim request is assigned a deep learning task of the requested task amount.
The non-volatile computer-readable storage medium of claim 17, wherein publishing the deep learning task on the blockchain platform includes:

Divide the deep learning task into multiple unit task volumes;

Calculate the price of unit task volume;

Determine the computing nodes that can apply for the deep learning task according to the computing environment required by the deep learning task;

The deep learning task is issued to a computing node that can apply for the deep learning task.
The non-volatile computer-readable storage medium according to claim 16, wherein performing deep learning on the deep learning task through the computing node to which it is distributed includes:

The computing node receives a deep learning task distributed to itself, where the deep learning task distributed to itself carries: a deep learning model and a training set;

The computing node inputs the training set as training samples into the deep learning model for training to obtain the trained deep learning model;

Wherein, the computing nodes in the blockchain platform include at least one of the following: full-featured nodes running on the FPGA server cluster, idle GPU computing nodes in enterprises, and idle GPU computing nodes in individuals.
The non-volatile computer-readable storage medium according to claim 16, wherein after performing deep learning on the deep learning task through the distributed computing node, the method further includes:

Acquiring execution information and completion status of the deep learning task;

Record the execution information and the completion status in the form of transactions in the blockchain platform.