Circuit search methods, training methods for search models, devices and electronic equipment

Abstract

This application provides a circuit search method, a search model training method, an apparatus, and an electronic device, belonging to the field of computer technology and relating to artificial intelligence and computer vision technology. When searching for a circuit, the user-inputted circuit name can be obtained. The circuit name, along with multiple circuit design diagrams and multiple circuit layout diagrams stored in a circuit diagram library, are input into a trained search model. The search model then outputs the circuit design diagram and circuit layout diagram corresponding to the circuit name. This circuit search method can quickly find and output the corresponding circuit design diagram and circuit layout diagram, saving time and improving work efficiency.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a circuit search method, a method for training a search model, an apparatus, and an electronic device. Background Technology

[0002] Electrical Failure Analysis (EFA) is a type of failure analysis primarily based on electrical measurements. Its main purpose is to identify the faulty test structures within a chip. When performing EFA, it is often necessary to first locate the circuit design diagram of the failed circuit in the layout software, and then find the corresponding circuit layout diagram.

[0003] Currently, when searching for a circuit, users need to do it manually. First, they need to open the circuit design schematic, find the circuit design diagram at the required location in the schematic, and then open the circuit layout schematic, manually finding the corresponding circuit layout diagram in the layout schematic. This process wastes users' time and reduces work efficiency.

[0004] Therefore, finding and outputting the corresponding circuit design diagrams and circuit layout diagrams is an urgent problem to be solved. Summary of the Invention

[0005] To address the problems in the prior art, embodiments of this application provide a method for training a search model, a circuit search method, an apparatus, and a device that can quickly find and output corresponding circuit design diagrams and circuit layout diagrams.

[0006] To achieve the above objectives, the technical solution of this application embodiment is implemented as follows:

[0007] In a first aspect, embodiments of this application provide a circuit search method, including:

[0008] Obtain the circuit name input by the user; input the circuit name, along with multiple circuit design diagrams and multiple circuit layout diagrams stored in the circuit diagram library, into the trained search model; output the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model.

[0009] In one optional embodiment, the step of outputting the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model includes:

[0010] The circuit name is converted into text feature information using the text recognition subnetwork of the search model; features are extracted from the multiple circuit design diagrams and multiple circuit layout diagrams using the image recognition subnetwork of the search model to obtain image feature information for each circuit design diagram and image feature information for each circuit layout diagram; based on the correlation between the text feature information and the image feature information of each circuit design diagram, the circuit design diagram corresponding to the circuit name is output; based on the correlation between the text feature information and the image feature information of each circuit layout diagram, the circuit layout diagram corresponding to the circuit name is output.

[0011] In an optional embodiment, the step of outputting the circuit design diagram corresponding to the circuit name based on the correlation between the text feature information and the image feature information of each circuit design diagram includes:

[0012] The correlation between the image feature information and the text feature information of each circuit design diagram is determined; the circuit design diagram with the highest correlation is taken as the circuit design diagram corresponding to the circuit name.

[0013] In an optional embodiment, the step of outputting the circuit design diagram corresponding to the circuit name based on the correlation between the text feature information and the image feature information of each circuit layout diagram includes:

[0014] The correlation between the image feature information and the text feature information of each circuit layout diagram is determined; the circuit layout diagram with the highest correlation is taken as the circuit layout diagram corresponding to the circuit name.

[0015] Secondly, embodiments of this application provide a method for training a search model, including:

[0016] Obtain a training sample set; the training sample set includes multiple circuit diagram sample pairs, each circuit diagram sample pair includes a circuit design sample diagram and a circuit layout sample diagram with a corresponding relationship; the circuit design sample diagram and the circuit layout sample diagram in each circuit diagram sample pair have the same circuit name label; extract circuit diagram sample pairs from the training sample set, and input the circuit design sample diagram and the circuit layout sample diagram contained in the extracted circuit diagram sample pair, as well as the circuit name label corresponding to the circuit design sample diagram and the circuit layout sample diagram, into the search model to be trained; determine the loss value according to the output of the search model to be trained and the correspondence between the circuit design sample diagram and the circuit layout sample diagram and the circuit name label; adjust the parameters of the search model to be trained according to the loss value until the loss value converges to the preset expected value, and obtain the trained search model.

[0017] In one optional embodiment, the search model includes an image recognition subnetwork and a text recognition subnetwork; determining the loss value based on the output of the search model to be trained and the correspondence between circuit design sample diagrams, circuit layout sample diagrams, and circuit name labels includes:

[0018] The image recognition sub-network of the search model extracts features from the extracted circuit diagram sample pairs, including circuit design sample diagrams and circuit layout sample diagrams, to obtain image feature information of the circuit design sample diagrams and image feature information of the circuit layout sample diagrams, respectively.

[0019] The text recognition subnetwork of the search model extracts features from the circuit name tag to obtain the text feature information of the circuit name tag; based on the image feature information and the text feature information, a first loss value is determined using a mode-invariant loss function; based on the image feature information and the text feature information, a second loss value is determined using a common space discriminant loss function; based on the image feature information and the text feature information, a third loss value is determined using a label space discriminant loss function; and based on the first loss value, the second loss value, and the third loss value, the final loss value is determined.

[0020] Thirdly, embodiments of this application provide a circuit search method apparatus, comprising:

[0021] The circuit name acquisition unit is used to acquire the circuit name input by the user.

[0022] The search unit is used to input the circuit name, as well as multiple circuit design diagrams and multiple circuit layout diagrams stored in the circuit diagram library, into the trained search model; and output the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model.

[0023] Fourthly, embodiments of this application provide a training apparatus for a search model, comprising:

[0024] A sample acquisition unit is used to acquire a training sample set; the training sample set includes multiple circuit diagram sample pairs, each circuit diagram sample pair includes a circuit design sample diagram and a circuit layout sample diagram with a corresponding relationship; the circuit design sample diagram and the circuit layout sample diagram in each circuit diagram sample pair have the same circuit name label;

[0025] The training unit is used to extract circuit diagram sample pairs from the training sample set, and input the circuit design sample diagrams and circuit layout sample diagrams, as well as the circuit name labels corresponding to the circuit design sample diagrams and circuit layout sample diagrams, into the search model to be trained; determine the loss value based on the output of the search model to be trained and the correspondence between the circuit design sample diagrams and circuit layout sample diagrams and the circuit name labels; and adjust the parameters of the search model to be trained according to the loss value until the loss value converges to a preset expected value, thereby obtaining the trained search model.

[0026] Fifthly, embodiments of this application provide an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, the processor enables the processor to implement the circuit search method of the first aspect or the training method of the search model of the second aspect.

[0027] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the circuit search method of the first aspect or the training method of the search model of the second aspect.

[0028] This application provides a circuit search method, a search model training method, an apparatus, and an electronic device. When searching for a circuit, the user-inputted circuit name can be obtained. The circuit name, along with multiple circuit design diagrams and multiple circuit layout diagrams stored in a circuit diagram library, are input into a trained search model. The search model then outputs the circuit design diagram and circuit layout diagram corresponding to the circuit name. This circuit search method allows for the rapid retrieval and output of corresponding circuit design diagrams and circuit layout diagrams, saving time and improving work efficiency. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A flowchart illustrating a training method for a search model provided in an embodiment of this application;

[0031] Figure 2 A schematic flowchart illustrating a circuit search method provided in an embodiment of this application;

[0032] Figure 3A schematic diagram illustrating the specific operation flow of a search model provided in an embodiment of this application;

[0033] Figure 4 This is a schematic diagram of the structure of a circuit search device provided in an embodiment of this application;

[0034] Figure 5 A schematic diagram of the structure of a training device for a search model provided in an embodiment of this application;

[0035] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0037] The word “exemplary” as used below means “serving as an example, embodiment, or illustration.” Any embodiment illustrated as an “exemplary” need not be construed as superior to or better than other embodiments.

[0038] The terms "first" and "second" used in this document are for descriptive purposes only and should not be construed as indicating relative importance or implying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0039] To help users quickly find corresponding circuit design diagrams and circuit layout diagrams, embodiments of this application provide a circuit search method, a search model training method, an apparatus, and a device. Embodiments of this application relate to Artificial Intelligence (AI) and machine learning technologies, and are designed based on Computer Vision (CV) and Machine Learning (ML) technologies within Artificial Intelligence.

[0040] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities. AI technology mainly includes several major areas such as computer vision, speech processing, and machine learning / deep learning.

[0041] With the research and advancement of artificial intelligence technology, AI is being studied and applied in various fields, such as smart homes, image retrieval, video surveillance, smart speakers, smart marketing, autonomous driving, drones, robots, and smart healthcare. It is believed that with the development of technology, AI will be applied in more fields and play an increasingly important role.

[0042] Computer vision technology is an important application of artificial intelligence. Research into its related theories and technologies aims to establish AI systems capable of extracting information from images, videos, or multidimensional data to replace human visual interpretation. Typical computer vision technologies usually include image processing and video analysis. The circuit search method provided in this application combines image processing and text processing methods.

[0043] Machine learning is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, and inductive learning. This application's embodiments employ a search model based on machine learning or deep learning. Based on the correlation between the textual feature information of the circuit name and the image feature information of each circuit diagram, it searches from a large number of circuit design diagrams and circuit layout diagrams to find the circuit design diagram and circuit layout diagram corresponding to the circuit name, and simultaneously outputs the circuit design diagram and circuit layout diagram.

[0044] The search model used in the embodiments of this application can be used Figure 1 The training method shown is used for training. This training method can be executed by an electronic device, which can be a server or a terminal device. Figure 1 As shown, the training method may include the following steps:

[0045] Step S101: Obtain the training sample set.

[0046] The server obtains a training sample set, which includes multiple circuit diagram sample pairs. Each circuit diagram sample pair includes a circuit design sample diagram and a circuit layout sample diagram with a corresponding relationship. The circuit design sample diagram and the circuit layout sample diagram in each circuit diagram sample pair have the same circuit name label. The training sample set can be created using a data annotation tool.

[0047] For example, the data annotation tool labellmg can be used to select the circuit design sample diagram and the corresponding circuit layout sample diagram to form a circuit diagram sample pair. The size range of the circuit design sample diagram and the corresponding circuit layout sample diagram can be selected according to actual needs. A circuit name label can be assigned to the selected circuit design sample diagram and the corresponding circuit layout sample diagram. For example, the circuit name label can be set to A.

[0048] In an optional embodiment, a large number of circuit diagram sample pairs at different locations, generated by the above method, and circuit name labels labeled on the circuit diagram sample pairs can be added to the training sample set to complete the generation of the training sample set.

[0049] Step S102: Extract circuit diagram sample pairs from the training sample set, and input the circuit design sample diagram and circuit layout sample diagram contained in the extracted circuit diagram sample pairs, as well as the circuit name labels corresponding to the circuit design sample diagram and circuit layout sample diagram, into the search model to be trained.

[0050] The server extracts circuit diagram sample pairs from the training sample set, and inputs the circuit design sample diagrams and circuit layout sample diagrams contained in the extracted circuit diagram sample pairs, as well as the circuit name labels corresponding to the circuit design sample diagrams and circuit layout sample diagrams, into the search model to be trained.

[0051] In one alternative embodiment, the search model may include an image recognition subnetwork and a text recognition subnetwork. The image recognition subnetwork is used to extract image feature information from circuit design sample diagrams and circuit layout sample diagrams, while the text recognition subnetwork is used to extract text feature information from circuit name labels.

[0052] For example, the image recognition sub-network can be implemented using Convolutional Neural Networks (CNNs), and its network structure can be, but is not limited to, VGGNet, ResNet, ShuffleNet, InceptionNet, etc. The image recognition sub-network can generate a 4096-dimensional feature vector as the original high-level semantic representation of the image using a VGGNet convolutional neural network, and then perform common representation learning to obtain image feature information. Compared to CNN networks, VGGNet includes more convolutional layers, and the convolutional kernels in each layer are smaller. This can be understood as VGGNet using multiple smaller convolutional kernels instead of one larger convolutional kernel in a CNN network. On the one hand, smaller kernels reduce the number of parameters; on the other hand, multiple convolutional kernels are equivalent to performing more non-linear mappings, which can increase the network's fitting ability.

[0053] The text recognition sub-network can be implemented using the Word2V-ec model and a text CNN network. The Word2V-ec model is used to obtain word vectors based on the input text. The word vectors are then input into the text CNN network to generate the original high-level semantic representation of the text. Common representation learning is then performed to obtain text feature information.

[0054] The extracted circuit diagram sample pairs, including the circuit design sample diagrams, are input into the image recognition sub-network of the search model to be trained, obtaining the image feature information of the circuit design sample diagrams output by the image recognition sub-network of the search model. Similarly, the extracted circuit diagram sample pairs, including the circuit layout sample diagrams, are input into the image recognition sub-network of the search model to be trained, obtaining the image feature information of the circuit layout sample diagrams output by the image recognition sub-network of the search model. Both the circuit design sample diagrams and the circuit layout sample diagrams can be referred to as circuit sample diagrams. The extracted circuit name tags are input into the text recognition sub-network of the search model to be trained, obtaining the text feature information of the circuit name tags output by the text recognition sub-network of the search model.

[0055] Step S103: Determine the loss value based on the output of the search model to be trained and the correspondence between the circuit design sample diagram, the circuit layout sample diagram and the circuit name label.

[0056] Using a predefined mode-invariant loss function, the mode-invariant loss value is determined based on the image and text feature information output by the search model to be trained. This mode-invariant loss value can be referred to as the first loss value μ1. The mode-invariant loss function can be expressed as:

[0057]

[0058] Where U represents the image feature information output by the search model, and V represents the text feature information output by the search model. Both image and text feature information can be understood as feature vectors, and UV can be understood as the difference matrix between the two feature vectors. F represents the L2 norm of the difference matrix (the square root of the sum of squares of corresponding elements), and n is the number of input circuit diagram sample pairs. When there is a correspondence between the circuit design sample diagrams and circuit layout sample diagrams input to the search model and the circuit name labels, the first loss value μ1 should approach 0; when there is no correspondence between the circuit design sample diagrams and circuit layout sample diagrams input to the search model and the circuit name labels, the first loss value μ1 should approach 1.

[0059] Using a predefined common space discriminative loss function, the common space discriminative loss value is determined based on the image and text feature information output by the search model to be trained. This common space discriminative loss value can be referred to as the second loss value μ2. The common space discriminative loss function can be expressed as:

[0060]

[0061] in, α represents the total number of input image samples, β represents the total number of input text samples, and cos(u i ,v j ) represents the cosine function, 1{u i ·v j} indicates a function, u i v represents the image feature information of the i-th circuit sample image. j This represents the textual feature information of the label of the j-th circuit sample.

[0062] Using a predefined label space discriminant loss function, the label space discriminant loss value is determined based on the image and text feature information output by the search model to be trained. This label space discriminant loss value can be referred to as the third loss value μ3. The label space discriminant loss function can be expressed as:

[0063]

[0064] Where U represents the image feature information output by the search model, V represents the text feature information output by the search model (both image and text feature information can be understood as feature vectors), T represents the T-th output image and text feature information, Y represents the one-hot representation of the circuit name label, and UV can be understood as the difference matrix between the two feature vectors. F represents the L2 norm of the difference matrix (the square root of the sum of squares of the corresponding elements), P represents the matrix projection of the linear classifier, and n is the number of input circuit diagram sample pairs.

[0065] Through the above process, the first loss value μ1, the second loss value μ2, and the third loss value μ3 are obtained respectively. Based on the first loss value, the second loss value, and the third loss value, the total loss value can be determined. The formula for calculating the total loss value can be expressed as: μ = μ1 + Aμ2 + Bμ3. Where A and B are hyperparameters that control the influence of the second loss value and the third loss value on the model. This total loss value is the loss value determined in step S103.

[0066] Step S104: Determine whether the loss value has converged to the preset expected value; if yes, proceed to step S106; if no, proceed to step S105.

[0067] The server determines whether the loss value has converged to the preset expected value, that is, whether it is less than or equal to the preset expected value. If it is, it means that the loss value has converged; if it is not, it means that the loss value has not yet converged.

[0068] Step S105: Adjust the parameters of the search model based on the loss value.

[0069] If the loss value has not yet converged, the server can use the backpropagation algorithm to adjust the parameters of the search model based on the loss value, and then return to step S101 to continue to extract training sample groups and continue to train the search model.

[0070] Step S106: Use the current parameters as parameters for the search model to obtain the trained search model.

[0071] The trained search model described above can be used for circuit search, finding the circuit design diagram and circuit layout diagram corresponding to the circuit name entered by the user. Figure 2 A flowchart illustrating an embodiment of this application also provides a circuit search method, which can be performed by an electronic device. Figure 2 As shown, the circuit search method may include the following steps:

[0072] Step S201: Obtain the circuit name input by the user.

[0073] During the circuit design process, users design circuit schematics and circuit layout schematics using circuit diagram design tools (such as Virtuoso). When the circuit schematics and circuit layout schematics cover a large area, users can select different locations on the circuit schematic to obtain multiple circuit schematics. Similarly, users can select different locations on the circuit layout schematic to obtain multiple circuit layout schematics. All the obtained circuit schematics and circuit layout schematics can be saved in the circuit library.

[0074] When a user wants to find the circuit design diagram and corresponding circuit layout diagram for a specific location, the user can enter the circuit name in the search box and click the search control. The electronic device receives the user's click on the search control and retrieves the circuit name entered by the user.

[0075] Step S202: Input the circuit name, as well as multiple circuit design diagrams and multiple circuit layout diagrams saved in the circuit diagram library, into the trained search model.

[0076] Step S203: Output the circuit design diagram and circuit layout diagram corresponding to the circuit name by searching the model.

[0077] The search model includes an image recognition subnetwork and a text recognition subnetwork. The image recognition subnetwork of the search model extracts features from multiple circuit design diagrams and multiple circuit layout diagrams, obtaining image feature information for each circuit design diagram and each circuit layout diagram, respectively. The text recognition subnetwork of the search model converts circuit names into text feature information.

[0078] Determine the correlation between the image feature information and text feature information of each circuit design diagram. The circuit design diagram with the highest correlation can be used as the circuit design diagram corresponding to the circuit name. Determine the correlation between the image feature information and text feature information of each circuit layout diagram. The circuit layout diagram with the highest correlation can be used as the circuit layout diagram corresponding to the circuit name. Output the circuit design diagram and the circuit layout diagram corresponding to the circuit name.

[0079] For example, a correlation function can be used to determine the correlation between image and text features of each circuit design diagram, and the correlation between image and text features of each circuit layout diagram. The correlation function can be a cosine function or a classification function. Image features can be understood as image feature vectors, and text features as text feature vectors. Using the correlation function, the cosine similarity between an image feature vector and a text feature vector can be determined. This cosine similarity can be used as the correlation between the image and text feature vectors. Alternatively, the cosine similarity can be converted into a probability value using the sigmoid function, and this probability value can be used as the correlation between the image and text feature vectors. A higher correlation indicates a greater probability that the image and text feature vectors correspond; a lower correlation indicates a lower probability. Therefore, the circuit design diagram with the highest correlation can be used as the circuit design diagram corresponding to the circuit name; and the circuit layout diagram with the highest correlation can be used as the circuit layout diagram corresponding to the circuit name.

[0080] In one alternative embodiment, the circuit name, along with any circuit design diagram stored in the circuit library, can be input into the trained system. Figure 3 The search model shown outputs the relevance between the circuit name and the circuit design diagram.

[0081] For example, assuming the user inputs the circuit name "operational amplifier," the text "operational amplifier" and any circuit design diagram saved in the circuit diagram library can be input into the trained search model. The text recognition subnetwork 302 of the search model converts the text "operational amplifier" into text feature information, and the image recognition subnetwork 301 of the search model extracts features from the input circuit design diagram to obtain the image feature information of the circuit design diagram. The obtained text feature information and image feature information are then input into a correlation function to obtain the relevance between the circuit name "operational amplifier" and the circuit design diagram. Similarly, the circuit name "operational amplifier" and any circuit layout diagram saved in the circuit diagram library can be input into the trained search model to obtain the relevance between the circuit name "operational amplifier" and the circuit layout diagram.

[0082] Assuming the circuit library contains 10 circuit design diagrams and 10 circuit layout diagrams, through... Figure 3 The search model shown can determine the relevance of 10 circuit design diagrams and 10 circuit layout diagrams respectively. For the 10 circuit design diagrams, the one with the highest relevance to the circuit name "operational amplifier" is selected as the circuit design diagram corresponding to the circuit name "operational amplifier". For the 10 circuit layout diagrams, the one with the highest relevance to the circuit name "operational amplifier" is selected as the circuit layout diagram corresponding to the circuit name "operational amplifier".

[0083] Corresponding to the embodiments of the circuit search method described above, this application also provides a circuit search device. Figure 4 This is a schematic diagram of the circuit search device provided in an embodiment of this application; as shown Figure 4 As shown, the circuit search device includes a circuit name acquisition unit 401 and a search unit 402. Wherein,

[0084] The circuit name acquisition unit 401 is used to acquire the circuit name input by the user;

[0085] The search unit 402 is used to input the circuit name, as well as multiple circuit design diagrams and multiple circuit layout diagrams stored in the circuit diagram library, into the trained search model; and output the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model.

[0086] In an optional embodiment, the search unit 402 is specifically used for:

[0087] The circuit name is converted into text feature information through the text recognition subnetwork of the search model; features are extracted from multiple circuit design diagrams and multiple circuit layout diagrams through the image recognition subnetwork of the search model, obtaining image feature information for each circuit design diagram and each circuit layout diagram respectively; based on the correlation between the text feature information and the image feature information of each circuit design diagram, the circuit design diagram corresponding to the circuit name is output; based on the correlation between the text feature information and the image feature information of each circuit layout diagram, the circuit layout diagram corresponding to the circuit name is output.

[0088] In an optional embodiment, the search unit 402 is specifically used for:

[0089] Determine the correlation between the image feature information and the text feature information of each circuit design diagram; select the circuit design diagram with the highest correlation as the circuit design diagram corresponding to the circuit name.

[0090] In an optional embodiment, the search unit 402 is specifically used for:

[0091] Determine the correlation between the image feature information and the text feature information of each circuit layout diagram; select the circuit layout diagram with the highest correlation as the circuit layout diagram corresponding to the circuit name.

[0092] Corresponding to the embodiments of the above-described search model training method, this application also provides a search model training apparatus, and the trained search model can be applied to the embodiments of the above-described circuit search method. Figure 5 This is a schematic diagram of the structure of the training device for the search model provided in this application embodiment; as shown Figure 5 As shown, the training device for this search model includes a sample acquisition unit 501 and a training unit 502. Wherein,

[0093] The sample acquisition unit 501 is used to acquire a training sample set; the training sample set includes multiple circuit diagram sample pairs, each circuit diagram sample pair includes a circuit design sample diagram and a circuit layout sample diagram with a corresponding relationship; the circuit design sample diagram and the circuit layout sample diagram in each circuit diagram sample pair have the same circuit name label;

[0094] Training unit 502 is used to extract circuit diagram sample pairs from the training sample set, and input the circuit design sample diagrams and circuit layout sample diagrams, as well as the circuit name labels corresponding to the circuit design sample diagrams and circuit layout sample diagrams, into the search model to be trained; determine the loss value based on the output of the search model to be trained and the correspondence between the circuit design sample diagrams and circuit layout sample diagrams and the circuit name labels; adjust the parameters of the search model to be trained according to the loss value until the loss value converges to the preset expected value, and obtain the trained search model.

[0095] In an optional embodiment, the training unit 502 is specifically used for:

[0096] The image recognition sub-network of the search model extracts features from the extracted circuit diagram samples, including circuit design sample diagrams and circuit layout sample diagrams, to obtain image feature information for the circuit design sample diagrams and circuit layout sample diagrams, respectively. The text recognition sub-network of the search model extracts features from the circuit name tags to obtain text feature information for the circuit name tags. Based on the image and text feature information, a modality-invariant loss function is used to determine the first loss value. Based on the image and text feature information, a common space discriminant loss function is used to determine the second loss value. Based on the image and text feature information, a label space discriminant loss function is used to determine the third loss value. The total loss value is determined based on the first, second, and third loss values. The parameters of the search model to be trained are adjusted according to the loss values ​​until the loss value converges to the preset expected value, thus obtaining the trained search model.

[0097] Corresponding to the above method embodiments, this application also provides an electronic device. This electronic device can be a server, a mobile terminal, or a computer, and includes at least a memory for storing data and a processor for data processing. The processor for data processing can be a microprocessor, CPU, GPU (Graphics Processing Unit), DSP, or FPGA. The memory stores operation instructions, which can be computer-executable code, to implement the various steps in the circuit search method or search model training method described in the above embodiments.

[0098] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; as shown below. Figure 6As shown, in this embodiment of the application, the electronic device 600 includes: a processor 601, a display 602, a memory 603, an input device 605, and a bus 604; the processor 601, the memory 603, the input device 605, and the display 602 are all connected through the bus 604, which is used for data transmission between the processor 601, the memory 603, the display 602, and the input device 605.

[0099] The memory 603 can be used to store software programs and modules, such as the program instructions / modules corresponding to the circuit search method or the training method of the search model in the embodiments of this application. The processor 601 executes various functional applications and data processing of the electronic device 600 by running the software programs and modules stored in the memory 603, such as the circuit search method provided in the embodiments of this application. The memory 603 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, the application program of at least one application, etc.; the data storage area may store data created according to the use of the electronic device 600 (such as the name of the circuit to be searched, the trained network model, and other related data). In addition, the memory 603 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0100] The processor 601 is the control center of the electronic device 600. It connects various parts of the electronic device 600 via a bus 604 and various interfaces and lines. It performs various functions and processes data of the electronic device 600 by running or executing software programs and / or modules stored in the memory 603, and by calling data stored in the memory 603. Optionally, the processor 601 may include one or more processing units, such as a CPU, GPU (Graphics Processing Unit), or digital processing unit.

[0101] In this embodiment, the processor 601 displays the results of the circuit search to the user via the display 602.

[0102] The input device 605 is mainly used to obtain user input operations, and the input device 605 may vary depending on the electronic device. For example, when the electronic device is a computer, the input device 605 may be a mouse, keyboard, or other input device 605. In this embodiment, the circuit name input by the user can be obtained through the input device.

[0103] This application also provides a computer storage medium storing computer-executable instructions for implementing the circuit search method described in any embodiment of this application.

[0104] This application also provides a computer storage medium storing computer-executable instructions for implementing the training method of the search model described in any embodiment of this application.

[0105] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A circuit search method, characterized in that, include: Get the circuit name input by the user; Input the circuit name, along with multiple circuit design diagrams and multiple circuit layout diagrams stored in the circuit diagram library, into the trained search model; The search model outputs the circuit design diagram and the circuit layout diagram corresponding to the circuit name. The step of outputting the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model includes: The circuit name is converted into text feature information through the text recognition sub-network of the search model; The image recognition sub-network of the search model is used to extract features from the multiple circuit design diagrams and the multiple circuit layout diagrams, respectively obtaining image feature information of each circuit design diagram and image feature information of each circuit layout diagram; Based on the correlation between the text feature information and the image feature information of each circuit design diagram, the circuit design diagram corresponding to the circuit name is output. Based on the correlation between the text feature information and the image feature information of each circuit layout diagram, the circuit layout diagram corresponding to the circuit name is output.

2. The method according to claim 1, characterized in that, The step of outputting the circuit design diagram corresponding to the circuit name based on the correlation between the text feature information and the image feature information of each circuit design diagram includes: The correlation between the image feature information and the text feature information of each circuit design diagram is determined respectively; The circuit design diagram with the highest relevance will be used as the circuit design diagram corresponding to the circuit name.

3. The method according to claim 1, characterized in that, The step of outputting the circuit design diagram corresponding to the circuit name based on the correlation between the text feature information and the image feature information of each circuit layout diagram includes: The correlation between the image feature information and the text feature information of each circuit layout diagram is determined respectively; The circuit layout diagram with the highest relevance will be used as the circuit layout diagram corresponding to the circuit name.

4. A method for training a search model, characterized in that, The training method includes: Obtain a training sample set; the training sample set includes multiple circuit diagram sample pairs, each circuit diagram sample pair includes a circuit design sample diagram and a circuit layout sample diagram with a corresponding relationship; the circuit design sample diagram and the circuit layout sample diagram in each circuit diagram sample pair have the same circuit name label; Circuit diagram sample pairs are extracted from the training sample set. The circuit design sample diagram and circuit layout sample diagram contained in the extracted circuit diagram sample pairs, as well as the circuit name labels corresponding to the circuit design sample diagram and circuit layout sample diagram, are input into the search model to be trained. The loss value is determined based on the output of the search model to be trained and the correspondence between the circuit design sample diagrams, the circuit layout sample diagrams and the circuit name labels. The parameters of the search model to be trained are adjusted according to the loss value until the loss value converges to the preset expected value, thus obtaining the trained search model.

5. The training method according to claim 4, characterized in that, The search model includes an image recognition subnetwork and a text recognition subnetwork; the step of determining the loss value based on the output of the search model to be trained and the correspondence between circuit design sample diagrams, circuit layout sample diagrams and circuit name labels includes: The image recognition sub-network of the search model extracts features from the extracted circuit diagram sample pairs, including circuit design sample diagrams and circuit layout sample diagrams, to obtain image feature information of the circuit design sample diagrams and image feature information of the circuit layout sample diagrams, respectively. The text recognition subnetwork of the search model is used to extract features from the circuit name tag to obtain the text feature information of the circuit name tag; Based on the image feature information and the text feature information, a first loss value is determined using a modality-invariant loss function; Based on the image feature information and the text feature information, a second loss value is determined using a common space discrimination loss function; Based on the image feature information and the text feature information, a third loss value is determined using a label space discriminant loss function; The loss value is determined based on the first loss value, the second loss value, and the third loss value.

6. A circuit search device, characterized in that, include: The circuit name acquisition unit is used to acquire the circuit name input by the user. The search unit is used to input the circuit name, as well as multiple circuit design diagrams and multiple circuit layout diagrams stored in the circuit diagram library, into the trained search model; and output the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model. The step of outputting the circuit design diagram and the circuit layout diagram corresponding to the circuit name through the search model includes: The circuit name is converted into text feature information through the text recognition sub-network of the search model; The image recognition sub-network of the search model is used to extract features from the multiple circuit design diagrams and the multiple circuit layout diagrams, respectively obtaining image feature information of each circuit design diagram and image feature information of each circuit layout diagram; Based on the correlation between the text feature information and the image feature information of each circuit design diagram, the circuit design diagram corresponding to the circuit name is output. Based on the correlation between the text feature information and the image feature information of each circuit layout diagram, the circuit layout diagram corresponding to the circuit name is output.

7. A training device for a search model, characterized in that, The training device includes: A sample acquisition unit is used to acquire a training sample set; the training sample set includes multiple circuit diagram sample pairs, each circuit diagram sample pair includes a circuit design sample diagram and a circuit layout sample diagram with a corresponding relationship; the circuit design sample diagram and the circuit layout sample diagram in each circuit diagram sample pair have the same circuit name label; The training unit is used to extract circuit diagram sample pairs from the training sample set, and input the circuit design sample diagrams and circuit layout sample diagrams, as well as the circuit name labels corresponding to the circuit design sample diagrams and circuit layout sample diagrams, into the search model to be trained; determine the loss value based on the output of the search model to be trained and the correspondence between the circuit design sample diagrams and circuit layout sample diagrams and the circuit name labels; adjust the parameters of the search model to be trained according to the loss value until the loss value converges to the preset expected value, thereby obtaining the trained search model.

8. An electronic device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, causes the processor to perform the method according to any one of claims 1 to 3 or any one of claims 4 to 5.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 3 or any one of claims 4 to 5.

Images