CN111882038A - Model conversion method and device - Google Patents
Model conversion method and device Download PDFInfo
- Publication number
- CN111882038A CN111882038A CN202010724863.9A CN202010724863A CN111882038A CN 111882038 A CN111882038 A CN 111882038A CN 202010724863 A CN202010724863 A CN 202010724863A CN 111882038 A CN111882038 A CN 111882038A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- processed
- neural network
- conversion
- 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
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 70
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000003062 neural network model Methods 0.000 claims abstract description 39
- 238000012545 processing Methods 0.000 claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000009466 transformation Effects 0.000 claims description 11
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000011426 transformation method Methods 0.000 claims description 6
- 238000009795 derivation Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 description 9
- 238000004590 computer program Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Document Processing Apparatus (AREA)
Abstract
The embodiment of the invention provides a model conversion method and a model conversion device. The method comprises the following steps: acquiring data to be processed from a data export interface of a source neural network model and determining a data arrangement format of the data to be processed; performing data arrangement format conversion on a model operator of a target neural network model according to the data arrangement format of the data to be processed; and processing the data to be processed by adopting the converted model operator. According to the method provided by the embodiment of the invention, the data arrangement format conversion is carried out on the model operator of the target neural network model, only one conversion operation needs to be executed, the operand is reduced, and the model conversion speed is increased.
Description
Technical Field
The invention relates to the technical field of computer science, in particular to a model conversion method and a model conversion device.
Background
Artificial intelligence has a wide application prospect in various industries. With the rapid development of artificial intelligence, various deep learning frameworks come into existence, and currently, mainstream deep learning frameworks based on neural networks include tensrflow, pytorreh, MXNet, and the like. In the process from algorithm development to algorithm deployment, different deep learning frameworks are often needed, for example, Pytorch is used for algorithm development and debugging, and in the deployment stage, due to the limitation of hardware and deployment software, the algorithm may need to be converted into a tensrflow model for adaptation. This requires inter-frame model conversion to achieve the goal of model adaptation hardware.
Because the data arrangement modes of different frames may be inconsistent, the data arrangement modes of the inter-frame model need to be converted in the conversion process. The existing model conversion method converts the data arrangement format of the source data into the target data arrangement format by performing conversion operation on the source data so as to realize model conversion between frames. However, the additional conversion operation increases the amount of computation, reducing the speed of model conversion.
Disclosure of Invention
The embodiment of the invention provides a model conversion method and a model conversion device, which are used for solving the problems of large computation amount and low speed of the existing model conversion method.
In a first aspect, an embodiment of the present invention provides a model conversion method, including:
acquiring data to be processed from a data export interface of a source neural network model and determining a data arrangement format of the data to be processed;
performing data arrangement format conversion on a model operator of the target neural network model according to the data arrangement format of the data to be processed;
and processing the data to be processed by adopting the converted model operator.
In one embodiment, the data arrangement format is NCHW, NHWC, or CHWN.
In one embodiment, the processing the data to be processed by using the transformed model operator includes:
and executing general matrix multiplication GEMM on the data to be processed and the converted model operator.
In one embodiment, obtaining the data to be processed from the data derivation interface of the source neural network model comprises:
and acquiring tensor data to be processed from the input end, the output end of the source neural network model, the input end of the hidden layer or the output end of the hidden layer.
In a second aspect, an embodiment of the present invention provides a model transformation apparatus, including:
the preprocessing module is used for acquiring data to be processed from a data export interface of the source neural network model and determining a data arrangement format of the data to be processed;
the conversion module is used for carrying out data arrangement format conversion on the model operator of the target neural network model according to the data arrangement format of the data to be processed;
and the processing module is used for processing the data to be processed by adopting the converted model operator.
In one embodiment, the data arrangement format is NCHW, NHWC, or CHWN.
In one embodiment, the processing module is configured to perform a generic matrix multiplication GEMM on the data to be processed and the transformed model operator.
In one embodiment, the preprocessing module is configured to obtain tensor data to be processed from an input end, an output end, an input end of the hidden layer, or an output end of the hidden layer of the source neural network model.
In a third aspect, an embodiment of the present invention provides a model conversion apparatus, including:
at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executes computer-executable instructions stored by the memory to cause the at least one processor to perform the model transformation method of any of the first aspects.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, the computer-readable storage medium is configured to implement the model transformation method according to any one of the first aspect.
The model conversion method and device provided by the embodiment of the invention comprise the following steps: acquiring data to be processed from a data export interface of a source neural network model and determining a data arrangement format of the data to be processed; performing data arrangement format conversion on a model operator of a target neural network model according to the data arrangement format of the data to be processed; and processing the data to be processed by adopting the converted model operator. By converting the data arrangement format of the model operator of the target neural network model, only one conversion operation needs to be executed, the operation amount is reduced, and the speed of model conversion is improved.
Drawings
FIG. 1 is a diagram illustrating a data arrangement format according to an embodiment;
FIG. 2 is a flow chart of an embodiment of a model transformation method provided by the present invention;
fig. 3 is a schematic diagram illustrating a model operator processing data to be processed according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of an embodiment of a model transformation apparatus according to the present invention;
fig. 5 is a schematic structural diagram of an embodiment of a model transformation apparatus provided in the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following detailed description and accompanying drawings. Wherein like elements in different embodiments are numbered with like associated elements. In the following description, numerous details are set forth in order to provide a better understanding of the present application. However, those skilled in the art will readily recognize that some of the features may be omitted or replaced with other elements, materials, methods in different instances. In some instances, certain operations related to the present application have not been shown or described in detail in order to avoid obscuring the core of the present application from excessive description, and it is not necessary for those skilled in the art to describe these operations in detail, so that they may be fully understood from the description in the specification and the general knowledge in the art.
Furthermore, the features, operations, or characteristics described in the specification may be combined in any suitable manner to form various embodiments. Also, the various steps or actions in the method descriptions may be transposed or transposed in order, as will be apparent to one of ordinary skill in the art. Thus, the various sequences in the specification and drawings are for the purpose of describing certain embodiments only and are not intended to imply a required sequence unless otherwise indicated where such sequence must be followed.
The numbering of the components as such, e.g., "first", "second", etc., is used herein only to distinguish the objects as described, and does not have any sequential or technical meaning. The term "connected" and "coupled" when used in this application, unless otherwise indicated, includes both direct and indirect connections (couplings).
Fig. 1 is a schematic diagram of a data arrangement format according to an embodiment. As shown in fig. 1, for the same data, different data arrangement formats are adopted, and the data in the memory are inconsistent. Wherein N represents the number, C represents the number of channels, H represents the height, and W represents the width. Taking image processing as an example, N may indicate how many images are in the batch, H may indicate how many pixels are in the vertical direction, W may indicate how many pixels are in the horizontal direction, and C may indicate the number of channels (for example, the number of channels C of a black-and-white image is 1, and the number of channels C of an RGB color image is 3).
NCHW is actually represented by [ wh cn ], the first element being 000, the second element being along W, i.e. 001, followed in turn by 002003; then along the H direction, 004005006007 … …; after 019, continue in direction C, wheel 020, after 021022 … …; up to 319 and then again in the N direction. Wherein [ a:0] is used to indicate a first dimension in the NCHW format, [ a:1] is used to indicate a second dimension in the NCHW format, [ a:2] is used to indicate a third dimension in the NCHW format, and [ a:3] is used to indicate a fourth dimension in the NCHW format.
NHWC in fact represents [ cwhn ], the first element being 000, the second along the C direction, i.e. 020, followed in turn by 040060 … …; up to 300 and then along W direction, 001021041061 … …; after 303, along the H direction, 004024 … …; up to 319 and then again in the N direction. Wherein [ b:0] is used to indicate a first dimension in the NHWC format, [ b:1] is used to indicate a second dimension in the NHWC format, [ b:2] is used to indicate a third dimension in the NHWC format, and [ b:3] is used to indicate a fourth dimension in the NHWC format.
The CHWN format is analogized, and is not described in detail herein.
Deep learning based on neural networks, such as deep convolutional neural networks, has become the most extensive network architecture, and has wide application in the fields of images, voice and the like. The core algorithm of the deep convolutional neural network is convolutional calculation, and the mainstream implementation manner at present is to expand the convolution and convert the convolutional calculation into General matrix multiplication (GEMM). GEMM requires that the data involved in the computation be in the same data arrangement format. In the process of model conversion, the source neural network model and the target neural network model may have different data arrangement formats. If the default data arrangement format of the tensrflow is NHWC and the default data arrangement format of ONNX is NCHW, the ONNX and the tensrflow data cannot be matched due to different data arrangement formats in the process of switching from the tensrflow to the ONNX or switching from the ONNX to the tensrflow, and if the parameters of the GEMM are not changed, the operation result is biased, thereby causing inference errors.
In order to avoid an error condition caused by a data arrangement format in a model conversion process, the existing model conversion method forcibly converts NHWC data of TensorFlow into NCHW data of ONNX by adding a conversion operator, or forcibly converts the NCHW data of ONNX into NHWC data of TensorFlow. However, in practice, the applicant finds that the data format conversion by adding the conversion operator increases the operation amount, and particularly when the data amount is large, such as the value of N is large, the speed of model conversion is greatly reduced.
Therefore, in order to reduce the operand and improve the speed of model conversion, the data arrangement format conversion is carried out on the model operator of the target neural network model instead of the data arrangement format conversion of the data of the source neural network model, so that only one conversion operation needs to be carried out no matter how large the data quantity needs to be processed in the model conversion process, the operand can be greatly reduced, and the speed of model conversion is improved.
Fig. 2 is a flowchart of a model transformation method according to an embodiment of the present invention. As shown in fig. 2, the model conversion method provided in this embodiment may include:
s201, obtaining data to be processed from a data export interface of the source neural network model and determining a data arrangement format of the data to be processed.
The data to be processed in this embodiment may be obtained from data export interfaces such as an input end, an output end, an input end of the hidden layer, or an output end of the hidden layer of the source neural network model according to the requirement of model conversion. The data arrangement format of the data to be processed includes, but is not limited to, NCHW, NHWC, or CHWN.
S202, converting the data arrangement format of the model operator of the target neural network model according to the data arrangement format of the data to be processed.
In this embodiment, after the data arrangement format of the data to be processed is determined, the data arrangement format of the model operator of the target neural network model is converted into a format that is the same as the data arrangement format of the data to be processed. For example, if the data arrangement format of the data to be processed is NHWC and the data arrangement format of the model operator of the target neural network model is NCHW, the model operator of the target neural network model is converted from NCHW to NHWC.
And S203, processing the data to be processed by adopting the converted model operator.
And after the data to be processed is consistent with the data arrangement format of the model operator, executing general matrix multiplication GEMM on the data to be processed and the converted model operator.
In order to improve the reliability of model conversion, the method provided by the embodiment may further perform consistency verification. For example, data for verification may be extracted from the data to be processed at a preset sampling frequency, and the extracted data may be subjected to data arrangement format conversion to be converted into a data arrangement format identical to the model operator of the target neural network model. Processing the verification data after the format conversion by adopting a model operator of the target neural network model, comparing a processing result with a result of processing the data to be processed by adopting the converted model operator, and if the processing result is the same as the result, indicating that the model conversion is accurate; if different, correction is required.
The following is a detailed description of how the method according to the embodiment of the present invention ensures the accuracy of the model transformation operation without adding operators. Referring to fig. 3, wherein C represents output data, a represents data to be processed, and B represents a model operator. M, N and K denote the dimensions of the data matrix. It should be noted that the size of each data matrix in the figure is only illustrated, and is not limited thereto.
The values of the elements C [ m ] [ n ] + ═ a [ m ] [ K ] × B [ K ] [ n ], K ═ 1, 2, … …, K in C. It is known that data in K dimension of a arranged in NCHW and NHWC may be inconsistent, for example, C512, H3, W3, and K4608. In the embodiment, the model operator is converted, so that the accuracy of the output data C is ensured. Assuming that the A matrix elements under the NCHW arrangement are A [ m ] [ k ], the B matrix elements are B [ k ] [ n ]; the A matrix elements in the NHWC arrangement are A [ m ] [ k ], and the B matrix elements are B [ k ] [ n ]. Given that the input C, H, W, 3 of the NCHW is 512, k, C, 9+ H, 3+ W in the NCHW, and k, H, 3+ W, 512+ C in the NHWC can be obtained. As long as k elements in the B matrix in NHWC and k elements in the B matrix in NCHW are guaranteed to be equal, consistency of C elements of output data can be guaranteed. I.e., a dimension conversion of (512, 3, 3) to (3, 3, 512) needs to be performed on the B matrix. In practical operation, 4608 dimensions in the B matrix can be firstly split into (512, 3, 3), then converted into (3, 3, 512), and finally spliced into 4608 dimensions again, so that the B matrix arranged under NHWC is generated. When the general matrix multiplication GEMM is executed on the data to be processed and the converted model operator, the conversion process does not influence the operation speed because the model operator is a constant. Through the steps, under the condition of ensuring the accuracy, the model conversion can be completed without adding a conversion operator, and compared with a processing mode of adding the conversion operator, the operation speed of the model conversion is improved.
In the model conversion method provided by this embodiment, to-be-processed data is acquired from a data export interface of a source neural network model, a data arrangement format of the to-be-processed data is determined, then, data arrangement format conversion is performed on a model operator of a target neural network model according to the data arrangement format of the to-be-processed data, and finally, the to-be-processed data is processed by using the converted model operator, so that model conversion is realized. By converting the data arrangement format of the model operator of the target neural network model, only one conversion operation needs to be executed, the operation amount is reduced, and the speed of model conversion is improved.
Fig. 4 is a schematic structural diagram of an embodiment of a model transformation apparatus provided in the present invention. As shown in fig. 4, the model transformation apparatus 40 provided in this embodiment may include: a pre-processing module 401, a conversion module 402 and a processing module 403.
The preprocessing module 401 is configured to obtain data to be processed from a data export interface of a source neural network model and determine a data arrangement format of the data to be processed;
a conversion module 402, configured to perform data arrangement format conversion on a model operator of a target neural network model according to a data arrangement format of the data to be processed;
and a processing module 403, configured to process the data to be processed by using the converted model operator.
The model transformation apparatus provided in this embodiment may be used to implement the technical solution of the method embodiment corresponding to fig. 2, and the implementation principle and the technical effect are similar, which are not described herein again.
In one embodiment, the data arrangement format is NCHW, NHWC, or CHWN.
In an embodiment, the processing module 403 is specifically configured to perform a generic matrix multiplication GEMM on the data to be processed and the transformed model operator.
In an embodiment, the preprocessing module 401 is specifically configured to obtain tensor data to be processed from an input end, an output end, an input end of the hidden layer, or an output end of the hidden layer of the source neural network model.
Fig. 5 shows a model transformation apparatus, which is only illustrated in fig. 5, and the embodiment of the present invention is not limited thereto. Fig. 5 is a schematic structural diagram of an embodiment of a model transformation apparatus provided in the present invention. As shown in fig. 5, the model conversion apparatus 50 provided in the present embodiment may include: memory 501, processor 502, and bus 503. The bus 503 is used to realize connection between the elements.
The memory 501 stores a computer program, and when the computer program is executed by the processor 502, the computer program can implement any of the technical solutions of the model conversion methods provided by the method embodiments.
Wherein, the memory 501 and the processor 502 are electrically connected directly or indirectly to realize the data transmission or interaction. For example, the components may be electrically connected to each other via one or more communication buses or signal lines, such as bus 503. The memory 501 stores a computer program for implementing the model conversion method, which includes at least one software functional module that can be stored in the memory 501 in the form of software or firmware, and the processor 502 executes various functional applications and data processing by running the software program and the module stored in the memory 501.
The Memory 501 may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like. The memory 501 is used for storing programs, and the processor 502 executes the programs after receiving execution instructions. Further, the software programs and modules within the memory 501 may also include an operating system, which may include various software components and/or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components.
The processor 502 may be an integrated circuit chip having signal processing capabilities. The Processor 502 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and so on. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. It will be appreciated that the configuration of fig. 5 is merely illustrative and may include more or fewer components than shown in fig. 5 or have a different configuration than shown in fig. 5. The components shown in fig. 5 may be implemented in hardware and/or software.
It should be noted that the model transformation device provided in this embodiment includes, but is not limited to, at least one of the following: user side equipment and network side equipment. User-side devices include, but are not limited to, computers, smart phones, tablets, digital broadcast terminals, messaging devices, game consoles, personal digital assistants, and the like. The network-side device includes, but is not limited to, a single network server, a server group consisting of a plurality of network servers, or a cloud consisting of a large number of computers or network servers based on cloud computing, wherein the cloud computing is one of distributed computing and is a super virtual computer consisting of a group of loosely coupled computers.
Reference is made herein to various exemplary embodiments. However, those skilled in the art will recognize that changes and modifications may be made to the exemplary embodiments without departing from the scope hereof. For example, the various operational steps, as well as the components used to perform the operational steps, may be implemented in differing ways depending upon the particular application or consideration of any number of cost functions associated with operation of the system (e.g., one or more steps may be deleted, modified or incorporated into other steps).
Additionally, as will be appreciated by one skilled in the art, the principles herein may be reflected in a computer program product on a computer readable storage medium, which is pre-loaded with computer readable program code. Any tangible, non-transitory computer-readable storage medium may be used, including magnetic storage devices (hard disks, floppy disks, etc.), optical storage devices (CD-ROMs, DVDs, Blu Ray disks, etc.), flash memory, and/or the like. These computer program instructions may be loaded onto a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions which execute on the computer or other programmable data processing apparatus create means for implementing the functions specified. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including means for implementing the function specified. The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified.
The present invention has been described in terms of specific examples, which are provided to aid understanding of the invention and are not intended to be limiting. For a person skilled in the art to which the invention pertains, several simple deductions, modifications or substitutions may be made according to the idea of the invention.
Claims (10)
1. A method of model conversion, comprising:
acquiring data to be processed from a data export interface of a source neural network model and determining a data arrangement format of the data to be processed;
performing data arrangement format conversion on a model operator of a target neural network model according to the data arrangement format of the data to be processed;
and processing the data to be processed by adopting the converted model operator.
2. The method of claim 1, wherein the data arrangement format is NCHW, NHWC, or CHWN.
3. The method of claim 1, wherein the processing the data to be processed using the transformed model operator comprises:
and executing general matrix multiplication GEMM on the data to be processed and the converted model operator.
4. The method of claim 1, wherein the obtaining the data to be processed from the data derivation interface of the source neural network model comprises:
and acquiring tensor data to be processed from the input end, the output end of the source neural network model, the input end of the hidden layer or the output end of the hidden layer.
5. A model conversion apparatus, characterized by comprising:
the preprocessing module is used for acquiring data to be processed from a data export interface of a source neural network model and determining a data arrangement format of the data to be processed;
the conversion module is used for carrying out data arrangement format conversion on the model operator of the target neural network model according to the data arrangement format of the data to be processed;
and the processing module is used for processing the data to be processed by adopting the converted model operator.
6. The apparatus of claim 5, wherein the data arrangement format is NCHW, NHWC, or CHWN.
7. The apparatus of claim 5, wherein the processing module is to perform a generalized matrix multiplication (GEMM) on the data to be processed and the transformed model operator.
8. The apparatus of claim 5, wherein the pre-processing module is configured to obtain tensor data to be processed from an input of a source neural network model, an output of the source neural network model, an input of a hidden layer, or an output of the hidden layer.
9. A model transformation apparatus, characterized by comprising: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the model transformation method of any of claims 1-4.
10. A computer-readable storage medium having computer-executable instructions stored thereon, which when executed by a processor, implement the model transformation method of any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010724863.9A CN111882038A (en) | 2020-07-24 | 2020-07-24 | Model conversion method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010724863.9A CN111882038A (en) | 2020-07-24 | 2020-07-24 | Model conversion method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111882038A true CN111882038A (en) | 2020-11-03 |
Family
ID=73200493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010724863.9A Pending CN111882038A (en) | 2020-07-24 | 2020-07-24 | Model conversion method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111882038A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113723601A (en) * | 2021-08-30 | 2021-11-30 | 北京市商汤科技开发有限公司 | Neural network model conversion method, device, equipment and storage medium |
| WO2022161060A1 (en) * | 2021-01-28 | 2022-08-04 | 展讯通信（上海）有限公司 | Data processing method and apparatus |
| CN114896950A (en) * | 2022-07-11 | 2022-08-12 | 浙江大华技术股份有限公司 | Model conversion method, model conversion device, and storage medium |
| CN116724316A (en) * | 2020-12-31 | 2023-09-08 | 华为技术有限公司 | Model processing method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110516707A (en) * | 2019-07-19 | 2019-11-29 | 深圳力维智联技术有限公司 | A kind of image labeling method and its device, storage medium |
| WO2020098414A1 (en) * | 2018-11-13 | 2020-05-22 | Oppo广东移动通信有限公司 | Data processing method for terminal, device, and terminal |
| CN111210005A (en) * | 2019-12-31 | 2020-05-29 | Oppo广东移动通信有限公司 | Equipment operation method and device, storage medium and electronic equipment |
| CN111401511A (en) * | 2019-09-24 | 2020-07-10 | 上海寒武纪信息科技有限公司 | Data processing method and device, computer equipment and storage medium |
-
2020
- 2020-07-24 CN CN202010724863.9A patent/CN111882038A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020098414A1 (en) * | 2018-11-13 | 2020-05-22 | Oppo广东移动通信有限公司 | Data processing method for terminal, device, and terminal |
| CN110516707A (en) * | 2019-07-19 | 2019-11-29 | 深圳力维智联技术有限公司 | A kind of image labeling method and its device, storage medium |
| CN111401511A (en) * | 2019-09-24 | 2020-07-10 | 上海寒武纪信息科技有限公司 | Data processing method and device, computer equipment and storage medium |
| CN111210005A (en) * | 2019-12-31 | 2020-05-29 | Oppo广东移动通信有限公司 | Equipment operation method and device, storage medium and electronic equipment |
Non-Patent Citations (2)
| Title |
|---|
| 元峰: "人脸口罩检测现开源PyTorch、TensorFlow、MXNet等五大主流深度学习框架模型和代码，及框架相互转换方法", pages 1 - 5, Retrieved from the Internet <URL:https://aijishu.com/a/1060000000093732> * |
| 谭喆: "物联网安全技术分析", 《中国安防》, no. 10, pages 67 - 73 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116724316A (en) * | 2020-12-31 | 2023-09-08 | 华为技术有限公司 | Model processing method and device |
| WO2022161060A1 (en) * | 2021-01-28 | 2022-08-04 | 展讯通信（上海）有限公司 | Data processing method and apparatus |
| CN113723601A (en) * | 2021-08-30 | 2021-11-30 | 北京市商汤科技开发有限公司 | Neural network model conversion method, device, equipment and storage medium |
| CN114896950A (en) * | 2022-07-11 | 2022-08-12 | 浙江大华技术股份有限公司 | Model conversion method, model conversion device, and storage medium |
| CN114896950B (en) * | 2022-07-11 | 2022-10-28 | 浙江大华技术股份有限公司 | Model conversion method, model conversion device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111882038A (en) | Model conversion method and device | |
| US20210365767A1 (en) | Method and device for operator registration processing based on deep learning and electronic device | |
| CN113065656A (en) | Rule engine configuration method and device, server and readable storage medium | |
| CN108170425B (en) | Program code modification method and device and terminal equipment | |
| CN113474760B (en) | Generating device, generating system, generating method, and recording medium | |
| CN114816393B (en) | Information generation method, device, equipment and storage medium | |
| CN109684008A (en) | Card rendering method, device, terminal and computer readable storage medium | |
| CN115150024B (en) | Data processing method, device, equipment and medium | |
| CN112631649A (en) | Intelligent contract management method, device, terminal equipment and medium | |
| CN111291443B (en) | Automatic generation method, device and equipment for I/O module model of vehicle control unit software | |
| CN109325167B (en) | Feature analysis method, device, equipment and computer-readable storage medium | |
| CN113626022A (en) | Object model creating method and device, storage medium and electronic equipment | |
| CN115940202A (en) | Multi-inverter power distribution control method, device and equipment based on artificial intelligence | |
| CN111258628B (en) | Rule file comparison method and device, readable storage medium and terminal equipment | |
| CN110764768A (en) | Method and device for mutual conversion between model object and JSON object | |
| CN111626401B (en) | Operation method and device | |
| CN111027196B (en) | Simulation analysis task processing method and device for power equipment and storage medium | |
| CN114047926A (en) | Method, device and equipment for converting array calculation expression language and storage medium | |
| CN109655072B (en) | Map generation method and device | |
| CN111338968A (en) | Project function module debugging method, device, medium and electronic equipment | |
| CN112286578A (en) | Method, apparatus, device and computer-readable storage medium executed by computing device | |
| CN113268230B (en) | Feedback information generation method and device, terminal and storage medium | |
| CN114492649B (en) | Model training method, device and equipment for deep learning model and storage medium | |
| CN117112490A (en) | Target model file generation method, value information generation method and device | |
| CN114490699A (en) | Electrical equipment matching method and device, control equipment 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201103 |
|
| RJ01 | Rejection of invention patent application after publication |