CN118428306A - A method and device for generating a probe card design scheme based on a multi-modal large model - Google Patents

Abstract

The present invention discloses a method and device for generating a probe card design scheme based on a multimodal large model, and relates to the field of chip testing technology. A specific implementation of the method includes: expanding a domain data set to obtain an extended data set; wherein the domain data set includes: probe card design specifications, and/or, effective probe card design schemes; pre-training based on a pre-training corpus, a domain data set, and an extended data set to obtain a basic model; wherein the basic model is a multimodal large model; based on an invalid probe layout scheme and/or a probe checkpoint log, a fine-tuning instruction data set is generated; based on the fine-tuning instruction data set, the basic model is fine-tuned to obtain a probe card design model. This implementation can save the cost of the probe card design scheme and improve design efficiency.

Description

A method and device for generating a probe card design scheme based on a multi-modal large model

Technical Field

The present invention relates to the field of chip testing technology, and in particular to a method and device for generating a probe card design solution based on a multi-modal large model.

Background technique

With the development of semiconductor technology, the design of probe cards in chip testing has become increasingly important. The design of probe cards can include the design of probe layout and probe parameter design. The design process not only needs to meet the requirements of electrical performance, but also needs to take into account the feasibility, cost and reliability of manufacturing. The traditional probe card design process relies on the experience of design engineers, which is not only time-consuming, but also difficult to meet the increasingly complex design requirements.

Summary of the invention

In view of this, an embodiment of the present invention provides a method and apparatus for generating a probe card design solution based on a multi-modal large model, which can save the cost of the probe card design solution and improve the design efficiency.

In a first aspect, an embodiment of the present invention provides a method for training a probe card design model, comprising:

Expanding the domain data set to obtain an extended data set; wherein the domain data set includes: a probe card design specification, and/or an effective probe card design solution;

Pre-training is performed based on the pre-training corpus, the domain dataset and the extended dataset to obtain a basic model; wherein the basic model is a multimodal large model;

generating a fine-tuning instruction dataset based on the invalid probe placement scheme and/or the probe checkpoint log;

The basic model is fine-tuned based on the fine-tuning instruction data set to obtain a probe card design model.

In a second aspect, an embodiment of the present invention provides a method for generating a probe card design solution based on a multi-modal large model, comprising:

Obtaining a probe card design instruction input by a user through an interactive interface;

Inputting the probe card design instruction into a probe card design model; wherein the probe card design model is trained based on the method described in any one of the above embodiments;

The probe card design solution output by the probe card design model is displayed on the interactive interface.

In a third aspect, an embodiment of the present invention provides a training device for a probe card design model, comprising:

An extension module is configured to extend the domain data set to obtain an extended data set; wherein the domain data set includes: a probe card design specification, and/or an effective probe card design solution;

A pre-training module, configured to perform pre-training based on a pre-training corpus, the domain dataset and the extended dataset to obtain a basic model; wherein the basic model is a multimodal large model;

An instruction generation module configured to generate a fine-tuning instruction dataset based on an invalid probe layout scheme and/or a probe checkpoint log;

The fine-tuning module is configured to fine-tune the basic model based on the fine-tuning instruction data set to obtain a probe card design model.

In a fourth aspect, an embodiment of the present invention provides a device for generating a probe card design solution based on a multi-modal large model, comprising:

An acquisition module configured to acquire a probe card design instruction input by a user through an interactive interface;

An input module configured to input the probe card design instruction into a probe card design model; wherein the probe card design model is obtained by training the device described in the above embodiment;

A display module is configured to display the probe card design solution output by the probe card design model on the interactive interface.

In a fifth aspect, an embodiment of the present invention provides an electronic device, including:

one or more processors;

a storage device for storing one or more programs,

When the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any of the above embodiments.

In a fifth aspect, an embodiment of the present invention provides a computer-readable medium having a computer program stored thereon, and when the program is executed by a processor, the method described in any of the above embodiments is implemented.

One embodiment of the above invention has the following advantages or beneficial effects: The embodiment of the present invention is based on a multimodal large model to train a probe card design model, which can assist users in probe card design, greatly shorten the design time, and allow users to quickly test multiple design solutions, thereby accelerating the development cycle of the probe card. The embodiment of the present invention uses accumulated probe card design specifications, valid probe card design solutions, invalid probe layout solutions, and probe checkpoint logs to train the probe card design model, which can improve the training quality of the probe card design model, thereby improving the quality of probe card design, reducing the dependence of the process on personal experience, and ensuring the consistency of design quality between different engineers.

The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

FIG1 is a flow chart of a method for training a probe card design model provided by an embodiment of the present invention;

FIG2 is a flow chart of a method for generating a probe card design solution based on a multi-modal large model provided by an embodiment of the present invention;

FIG3 is a flow chart of a method for generating a probe card design solution based on a multi-modal large model provided by another embodiment of the present invention;

FIG4 is a schematic diagram of an interactive interface provided by an embodiment of the present invention;

FIG5 is a schematic diagram of a probe card design model training device provided by an embodiment of the present invention;

FIG6 is a schematic diagram of an apparatus for generating a probe card design solution based on a multi-modal large model provided by an embodiment of the present invention;

FIG. 7 is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention.

Detailed ways

The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

As shown in FIG1 , an embodiment of the present invention provides a method for training a probe card design model, including:

Step 101: Expand a domain data set to obtain an extended data set; wherein the domain data set includes: probe card design specifications, and/or effective probe card design solutions.

The design of the probe card can include two aspects: one is the design of probe parameters, such as probe material, probe length, etc., and the other is the design of probe layout, such as the interval between probes, etc. In actual application scenarios, only one of the above aspects can be included, or both aspects can be included at the same time.

Domain datasets refer to datasets related to probe card design, which focus on specific fields compared to pre-trained corpora. Probe card design specifications refer to the design standards that need to be followed in the probe card design process, and their content can exist in the form of pictures, text, etc. An effective probe card design scheme refers to an available probe card design scheme, that is, a probe card design scheme that meets the design requirements. For example, there are existing cases of probe card layout design for different signal categories (such as SIE, SATA, USB, PCIE), which consist of text and pictures; there are existing cases of probe parameter design for different machine types. If the probe card design scheme does not meet the design requirements, it is an invalid probe card design scheme, such as if its probe length does not meet the probe card design specifications.

Corresponding to the above two aspects, the probe card design specification may include a probe layout design specification and/or a parameter design specification. Similarly, an effective probe card design solution includes an effective probe layout design solution and/or an effective probe parameter design solution. The design probe design specification may be specified by an engineer according to business requirements.

Considering that the number of domain data sets is limited, the embodiment of the present invention generates more domain data based on the existing domain data sets to ensure the quality of pre-training.

Step 102: Pre-training is performed based on the pre-training corpus, the domain dataset, and the extended dataset to obtain a basic model; wherein the basic model is a multimodal large model.

The pre-training corpus is a general corpus, such as MNBVC (Massive Never-ending BT Vast Chinese Corpus), WuDaoCorporaText, etc.

The basic model belongs to the transformer model. In order to be suitable for processing multimodal data such as pictures and texts, the embodiment of the present invention adopts a multimodal large model.

Step 103: Generate a fine-tuning instruction dataset based on the invalid probe placement solution and/or the probe checkpoint log.

An invalid probe layout plan refers to a probe layout plan that does not meet design requirements, such as error cases accumulated by engineers during the design process. An invalid probe layout plan may include invalid design pictures, invalid description information, correction suggestions, and other content.

A probe checkpoint log is a log file that records the behavior of the probe at key checkpoints in network and system activities. It usually contains detailed information about how the probe detects, analyzes, and responds to various events and activities.

Step 104: Fine-tune the basic model based on the fine-tuning instruction data set to obtain a probe card design model.

By fine-tuning the base model, the base model can learn more information about the probe card design, thereby improving its accuracy and generalization ability on the probe card design.

The embodiment of the present invention is based on a multimodal large model to train a probe card design model, which can assist users in designing probe cards, greatly shorten the design time, and allow users to quickly test multiple design solutions, thereby accelerating the development cycle of the probe card. The embodiment of the present invention uses accumulated probe card design specifications, effective probe card design solutions, invalid probe layout solutions, and probe checkpoint logs to train the probe card design model, which can improve the training quality of the probe card design model, thereby improving the quality of probe card design, reducing the dependence of the process on personal experience, and ensuring the consistency of design quality between different engineers.

In one embodiment of the present invention, the domain data set is a collection of text and image pairs, and the domain data includes any one or more parameters of probe height, electrical contact point spacing, probe alignment accuracy, signal integrity, power integrity, grain size and grain spacing.

Among them, the probe card design specifications and the effective probe card design plan can both contain text and pictures. The probe height, electrical contact point spacing and probe alignment accuracy are information related to the probe parameter design, which can be derived from the probe card design specifications or the effective probe card design plan. Signal integrity, power integrity, grain size and grain spacing are information related to the probe layout design, which can be derived from the probe card design specifications or the effective probe card design plan.

For the parameters in the domain data, their interpretations can refer to Table 1, where the design requirements can be determined by the probe card design specifications, and can also be determined by the user based on actual scenario requirements.

By pre-training with the various parameters shown in Table 1, the training quality of the basic model can be improved, making the probe card design model suitable for a variety of application scenarios.

Table 1 Explanation of parameters in domain data

In one embodiment of the present invention, the domain data set is expanded to obtain an extended data set, including:

Based on the design specification of the probe card, description prompt words and description seeds are generated, and the description prompt words and description seeds are input into the general large model to obtain a description data set;

Based on the effective probe card design, generate question-answer prompt words and question-answer seeds, input the prompt words and question-answer seeds into the general large model, and obtain the question-answer pair dataset;

Based on the probe card design specification and the effective probe card design scheme, a design suggestion prompt word and a design suggestion seed are generated, and the design suggestion prompt word and the design suggestion seed are input into the general macro model to obtain a design suggestion data set;

Among them, the description dataset, question-answer pair dataset, and design suggestion dataset constitute the extended dataset.

The general large model is an existing multimodal large model, which can be a GPT (Generative Pre-Trained) series model, such as GPT-4, or BERT (Bidirectional Encoder Representations from Transformers).

Examples of prompt words and seeds for the above three datasets are shown in Table 2, where Prompt is used to represent prompt words, and Seed is used to represent seeds. Different types of datasets have corresponding prompt word and seed pairs.

Table 2. Examples of prompt words and seeds in the dataset

Through the above three prompt words and seed pairs, the pre-trained data set can be expanded from multiple dimensions to obtain a richer extended data set and improve the training quality.

In one embodiment of the present invention, generating a fine-tuning instruction data set based on an invalid probe layout scheme and/or a probe checkpoint log includes:

extracting an original layout plan, a revised plan description and a revised layout plan from the invalid probe layout plan;

Extract the probe parameters before optimization, the optimization scheme description, and the probe parameters after optimization of the probe from the probe checkpoint log;

Generate layout correction hint words and layout correction seeds based on the original layout plan, the correction plan description and the correction layout plan;

Generate parameter optimization prompt words and parameter optimization seeds based on the probe parameters before optimization, the optimization scheme description and the probe parameters after optimization;

Inputting the layout correction prompt word and the layout correction seed into the universal large model to obtain the layout correction instruction;

Input parameter optimization prompt words and parameter optimization seeds into the general large model to obtain parameter optimization instructions;

Among them, layout correction instructions and parameter optimization instructions constitute the fine-tuning instruction data set.

The general large model can be GPT, BERT, etc.

For example, the original layout plan is the original design drawing, the revised plan description is the cause of the error and the revised plan description, and the revised layout plan is the revised design drawing.

The prompt words and seed examples of the above two instructions are shown in Table 3. Constructing instructions based on invalid probe layout solutions and probe checkpoint logs can enable the model to analyze problems existing in previous probe card designs and obtain higher quality probe card design solutions.

Table 3. Example of prompt words and seeds for commands

As shown in FIG. 2 , an embodiment of the present invention provides a method for generating a probe card design solution based on a multi-modal large model, including:

Step 201: Acquire a probe card design instruction input by a user through an interactive interface.

The probe card design instruction may be in the form of text, voice, etc., such as inputting the text "design a probe card using copper probes with a needle length of 2.5 mm, a needle width of 0.1 mm, and 20 probes on the probe card".

Step 202: Input the probe card design instruction into the probe card design model; wherein the probe card design model is trained based on the above-mentioned probe card design model training method.

Step 203: Display the probe card design solution output by the probe card design model on the interactive interface.

Users can use the probe card design model to design a probe card design that meets business requirements.

In one embodiment of the present invention, the method further comprises:

Generate vectors of multiple reference texts; wherein the multiple reference texts include: any one or more of a probe card design specification, a valid probe card design solution, an invalid probe layout solution, and a probe checkpoint log;

Obtaining the query command input by the user through the interactive interface;

Determine the similarity between the query vector generated by the query instruction and the vectors of each reference text;

Display the target reference text in the interactive interface based on the order of similarity from high to low;

The step of obtaining the probe card design instruction input by the user through the interactive interface includes:

A probe card design instruction for a target reference text input by a user through an interactive interface is obtained.

In order to improve the design quality of the probe card, the embodiment of the present invention constructs a vector of reference text based on any one or more of the probe card design specifications, effective probe card design solutions, invalid probe layout solutions, and probe checkpoint logs, so as to provide prompts for the probe card design specifications and common problems in the design during the user design process. For example, the query instruction is the text "probe material design", and the interactive interface displays "metal or metal alloy". Of course, the interactive interface can display the target reference text or keywords generated by the target reference text.

In one embodiment of the present invention, the method further comprises:

Based on the probe card design solution, domain data is generated and added to a domain data set; wherein the domain data set is used to train the probe card design model.

The embodiment of the present invention adds the probe card design scheme obtained through the probe card design model to the domain data set to further adjust the probe card design model and improve the training quality of the probe card design model.

In one embodiment of the present invention, the method further comprises:

Based on the query instructions and the probe card design instructions, fine-tuning instruction data is generated, and the fine-tuning instruction data is added to a fine-tuning instruction data set; wherein the fine-tuning instruction data set is used to train the probe card design model.

Specifically, the above-mentioned general large model can be used to analyze the query instructions and probe card design instructions, extract the characteristics such as probe parameters and performance indicators, and obtain the fine-tuning instruction data. In other words, the fine-tuning instruction data can be obtained based on the invalid probe layout plan and/or the probe checkpoint log, and can also be obtained based on the query instructions and the probe card design instructions. The embodiment of the present invention expands the fine-tuning instruction data set based on the query instructions and the probe card design instructions to further improve the quality of model training.

As shown in FIG3 , an embodiment of the present invention provides a method for generating a probe card design solution based on a multi-modal large model, including:

Step 301: Generate description prompt words and description seeds based on the probe card design specification, input the description prompt words and description seeds into the general macro model, and obtain a description data set.

Step 302: Based on the effective probe card design solution, generate question and answer prompt words and question and answer seeds, input the prompt words and question and answer seeds into the general large model, and obtain a question and answer pair data set.

Step 303: Based on the probe card design specification and the effective probe card design solution, generate design suggestion prompt words and design suggestion seeds, input the design suggestion prompt words and design suggestion seeds into the general macro model, and obtain a design suggestion data set.

Step 304: Pre-training is performed based on the pre-training corpus, the extended dataset and the domain dataset to obtain a basic model; wherein the basic model is a multimodal large model, and the extended dataset includes a description dataset, a question-answer pair dataset and a design suggestion dataset.

Step 305: extract the original layout plan, the revised plan description and the revised layout plan from the invalid probe layout plan, and extract the probe parameters before optimization, the optimized plan description and the probe parameters after optimization from the probe checkpoint log.

Step 306: Generate layout modification hint words and layout modification seeds based on the original layout plan, the modified plan description and the modified layout plan; generate parameter optimization hint words and parameter optimization seeds based on the probe parameters before optimization, the optimization plan description and the probe parameters after optimization.

Step 307: Input the layout correction prompt word and the layout correction seed into the universal macro model to obtain the layout correction instruction, and input the parameter optimization prompt word and the parameter optimization seed into the universal macro model to obtain the parameter optimization instruction.

Step 308: Fine-tune the basic model based on the fine-tuning instruction data set to obtain a probe card design model. The fine-tuning instruction data set includes layout correction instructions and parameter optimization instructions.

Step 309: Generate vectors of multiple reference texts.

Step 310: Obtain a query instruction input by a user through an interactive interface, and determine the similarity between a query vector generated by the query instruction and a vector of each reference text.

Step 311: Display the target reference text in the interactive interface based on the order of similarity from high to low.

Step 312: Obtaining probe card design instructions for the target reference text input by the user through the interactive interface.

Step 313: Input the probe card design instructions into the probe card design model.

Step 314: Display the probe card design solution output by the probe card design model on the interactive interface.

The method may also include converting the probe card design in a picture format into a vector graphic.

As shown in Figure 4, the interactive interface includes a reference tracking area, a Chatbot area, a converter area, and a feedback area. Among them, the reference tracking area is used to display information such as probe card design specifications, valid probe card design plans, invalid probe layout plans, and probe checkpoint logs and their source documents. The Chatbot area is used to display conversations, such as user query instructions, probe card design instructions, and probe card design plans. The converter area is used to convert the design drawings in image format into editable vector graphics for use in simulation software, which is convenient for engineers to verify the probe card design plan. In the feedback area, users can submit the final selected probe card design plan and related instructions to the feedback area to facilitate iteration of the probe card design model.

The embodiment of the present invention fully considers the complexity and diversity of probe card design, can efficiently process and analyze multimodal data, and generate high-quality probe card design solutions. During the model training process, the probe card design specifications, effective probe card design solutions, invalid probe layout solutions, and probe checkpoint logs are considered, which can improve the quality of model training and output higher-quality probe card design solutions.

As shown in FIG5 , an embodiment of the present invention provides a training device for a probe card design model, including:

The expansion module 501 is configured to expand the domain data set to obtain an extended data set; wherein the domain data set includes: a probe card design specification, and/or an effective probe card design solution;

A pre-training module 502 is configured to perform pre-training based on a pre-training corpus, a domain data set, and an extended data set to obtain a basic model; wherein the basic model is a multi-modal large model;

An instruction generation module 503 is configured to generate a fine-tuning instruction data set based on the invalid probe layout scheme and/or the probe checkpoint log;

The fine-tuning module 504 is configured to fine-tune the basic model based on the fine-tuning instruction data set to obtain a probe card design model.

In one embodiment of the present invention, the domain data set is a collection of text and image pairs, and the domain data includes any one or more parameters of probe height, electrical contact point spacing, probe alignment accuracy, signal integrity, power integrity, grain size and grain spacing.

In one embodiment of the present invention, the extension module 501 is configured to generate description prompt words and description seeds based on the probe card design specification, input the description prompt words and description seeds into the general macro model to obtain a description data set; generate question and answer prompt words and question and answer seeds based on the effective probe card design plan, input the prompt words and question and answer seeds into the general macro model to obtain a question and answer pair data set; generate design suggestion prompt words and design suggestion seeds based on the probe card design specification and the effective probe card design plan, input the design suggestion prompt words and design suggestion seeds into the general macro model to obtain a design suggestion data set; wherein the description data set, the question and answer pair data set and the design suggestion data set constitute an extended data set.

In one embodiment of the present invention, the instruction generation module 503 is configured to extract the original layout plan, the correction plan description and the correction layout plan from the invalid probe layout plan; extract the probe parameters before optimization, the optimization plan description and the probe parameters after optimization of the probe from the probe checkpoint log; generate layout correction prompt words and layout correction seeds based on the original layout plan, the correction plan description and the correction layout plan; generate parameter optimization prompt words and parameter optimization seeds based on the probe parameters before optimization, the optimization plan description and the probe parameters after optimization; input the layout correction prompt words and the layout correction seeds into the general large model to obtain layout correction instructions; input the parameter optimization prompt words and the parameter optimization seeds into the general large model to obtain parameter optimization instructions; wherein, the layout correction instructions and the parameter optimization instructions constitute a fine-tuning instruction data set.

As shown in FIG6 , an embodiment of the present invention provides a device for generating a probe card design solution based on a multi-modal large model, including:

An acquisition module 601 is configured to acquire a probe card design instruction input by a user through an interactive interface;

An input module 602 is configured to input a probe card design instruction into a probe card design model; wherein the probe card design model is obtained by training based on a training device for the probe card design model;

The display module 603 is configured to display the probe card design solution output by the probe card design model on the interactive interface.

In one embodiment of the present invention, an acquisition module 601 is configured to generate vectors of multiple reference texts; wherein the multiple reference texts include: any one or more of: probe card design specifications, valid probe card design plans, invalid probe layout plans and probe checkpoint logs; obtain query instructions input by the user through an interactive interface; determine the similarity between the query vector generated by the query instruction and the vectors of each reference text; the display module 603 is configured to display the target reference text on the interactive interface in descending order of similarity; the acquisition module 601 is configured to obtain the probe card design instructions for the target reference text input by the user through the interactive interface.

In one embodiment of the present invention, the acquisition module 601 is configured to generate domain data based on the probe card design solution, and add the domain data to the domain data set; wherein the domain data set is used to train the probe card design model.

In one embodiment of the present invention, the acquisition module 601 is configured to generate fine-tuning instruction data based on the query instruction and the probe card design instruction, and add the fine-tuning instruction data to the fine-tuning instruction data set; wherein the fine-tuning instruction data set is used to train the probe card design model.

Referring to Figure 7, a schematic diagram of a computer system 700 suitable for implementing a terminal device of an embodiment of the present invention is shown. The terminal device shown in Figure 7 is only an example and should not limit the functions and scope of use of the embodiment of the present invention.

As shown in FIG7 , the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. Various programs and data required for the operation of the system 700 are also stored in the RAM 703. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to the bus 704.

The following components are connected to the I/O interface 705: an input section 706 including a keyboard, a mouse, etc.; an output section 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, a modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 710 as needed, so that a computer program read therefrom is installed into the storage section 708 as needed.

In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 709, and/or installed from the removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, the above-mentioned functions defined in the system of the present invention are executed.

It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor includes a sending module, an acquisition module, a determination module, and a first processing module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the sending module may also be described as a "module for sending a picture acquisition request to the connected server."

The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims (10)

1. A method for training a probe card design model, comprising: Expanding the domain data set to obtain an extended data set; wherein the domain data set includes: a probe card design specification, and/or an effective probe card design solution; Pre-training is performed based on the pre-training corpus, the domain dataset and the extended dataset to obtain a basic model; wherein the basic model is a multimodal large model; generating a fine-tuning instruction dataset based on the invalid probe placement scheme and/or the probe checkpoint log; The basic model is fine-tuned based on the fine-tuning instruction data set to obtain a probe card design model. 2. The method according to claim 1, characterized in that The domain data set is a collection of text and picture pairs, and the domain data includes any one or more parameters of probe height, electrical contact point spacing, probe alignment accuracy, signal integrity, power integrity, grain size, and grain spacing. 3. The method according to any one of claims 1 or 2, characterized in that: The domain dataset is expanded to obtain an extended dataset, including: Based on the probe card design specification, generating description prompt words and description seeds, inputting the description prompt words and the description seeds into a universal macro model to obtain a description data set; Based on the effective probe card design, generating question-answer prompt words and question-answer seeds, inputting the prompt words and the question-answer seeds into the universal macro model, and obtaining a question-answer pair data set; Based on the probe card design specification and the effective probe card design solution, generating a design suggestion prompt word and a design suggestion seed, inputting the design suggestion prompt word and the design suggestion seed into the general macro model to obtain a design suggestion data set; Among them, the description dataset, the question-answer pair dataset and the design suggestion dataset constitute the extended dataset. 4. The method according to claim 1, characterized in that Generate a fine-tuning instruction dataset based on the invalid probe placement scheme and/or probe checkpoint log, including: extracting an original layout plan, a revised plan description and a revised layout plan from the invalid probe layout plan; Extracting the pre-optimization probe parameters, the optimization scheme description, and the post-optimization probe parameters of the probe from the probe checkpoint log; Based on the original layout scheme, the revised scheme description and the revised layout scheme, generating a layout revision prompt word and a layout revision seed; Generate parameter optimization prompt words and parameter optimization seeds based on the pre-optimization probe parameters, the optimization scheme description and the post-optimization probe parameters; Inputting the layout correction prompt word and the layout correction seed into a universal macro model to obtain a layout correction instruction; Inputting the parameter optimization prompt word and the parameter optimization seed into the universal large model to obtain a parameter optimization instruction; The layout correction instructions and the parameter optimization instructions constitute the fine-tuning instruction data set. 5. A method for generating a probe card design based on a multi-modal large model, characterized by comprising: Obtaining a probe card design instruction input by a user through an interactive interface; Inputting the probe card design instruction into a probe card design model; wherein the probe card design model is trained based on the method according to any one of claims 1 to 4; The probe card design solution output by the probe card design model is displayed on the interactive interface. 6. The method according to claim 5, further comprising: Generate vectors of multiple reference texts; wherein the multiple reference texts include: any one or more of a probe card design specification, a valid probe card design solution, an invalid probe layout solution, and a probe checkpoint log; Obtaining a query instruction input by a user through the interactive interface; Determining the similarity between the query vector generated by the query instruction and the vectors of each of the reference texts; Based on the order of the similarities from high to low, displaying the target reference text on the interactive interface; The step of obtaining the probe card design instruction input by the user through the interactive interface includes: Acquire a probe card design instruction for the target reference text input by the user through the interactive interface. 7. The method of claim 5, further comprising: Based on the probe card design solution, domain data is generated, and the domain data is added to a domain data set; wherein the domain data set is used to train the probe card design model. 8. The method of claim 5, further comprising: Based on the query instruction and the probe card design instruction, fine-tuning instruction data is generated, and the fine-tuning instruction data is added to a fine-tuning instruction data set; wherein the fine-tuning instruction data set is used to train the probe card design model. 9. A probe card design model training device, comprising: An extension module is configured to extend the domain data set to obtain an extended data set; wherein the domain data set includes: a probe card design specification, and/or an effective probe card design solution; A pre-training module, configured to perform pre-training based on a pre-training corpus, the domain dataset and the extended dataset to obtain a basic model; wherein the basic model is a multimodal large model; An instruction generation module configured to generate a fine-tuning instruction dataset based on an invalid probe layout scheme and/or a probe checkpoint log; The fine-tuning module is configured to fine-tune the basic model based on the fine-tuning instruction data set to obtain a probe card design model. 10. A device for generating a probe card design scheme based on a multi-modal large model, characterized by comprising: An acquisition module configured to acquire a probe card design instruction input by a user through an interactive interface; An input module configured to input the probe card design instruction into a probe card design model; wherein the probe card design model is obtained by training based on the apparatus according to claim 9; A display module is configured to display the probe card design solution output by the probe card design model on the interactive interface.