Flash memory characteristic prediction model training method, flash memory characteristic prediction method and device

Abstract

The application provides a flash memory characteristic prediction model training method, a flash memory characteristic prediction method and device, and relates to the technical field of artificial intelligence, which comprises the following steps: constructing graph structure data; the graph structure data comprises multiple nodes and edges between the nodes; the nodes are determined based on first target information of at least one type of flash memory; the first target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information; the edges represent the interaction relationship between the nodes; and an initial flash memory characteristic prediction model is trained based on the graph structure data to obtain a target flash memory characteristic prediction model. The graph structure data is constructed by using the multi-dimensional target information of various types of flash memory and the interaction relationship therebetween, and the target flash memory characteristic prediction model obtained by training can realize accurate prediction of flash memory characteristics and improve the accuracy of flash memory characteristic prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a flash memory characteristic prediction model training method and a flash memory characteristic prediction method and device. BACKGROUND

[0002] With the wide application of solid state drives (SSDs), flash memory (NAND Flash) as its core storage medium, its performance directly affects the overall performance of the SSD. However, with the continuous progress of NAND flash memory technology, the physical size of the storage unit is continuously reduced, making the reliability problem of NAND flash memory increasingly prominent. Therefore, how to accurately predict the characteristics of NAND flash memory and then optimize the design and use strategy of the storage system has become one of the current research hotspots.

[0003] Traditional NAND flash memory characteristic prediction methods mainly rely on empirical models and simplified mathematical models. Although these methods can reflect the characteristics of NAND flash memory to some extent, they cannot fully consider complex factors in actual working conditions, such as temperature, voltage fluctuations, usage patterns, etc., so the prediction accuracy is limited. In addition, with the continuous development of NAND flash memory technology, the limitations of traditional methods are increasingly prominent, and cannot meet the growing data storage demand.

[0004] In the prior art, some methods based on machine learning algorithms have been proposed to predict the characteristics of NAND flash memory. These methods mainly use neural networks, support vector machines, random forests, and other machine learning models to model and predict key parameters of NAND flash memory such as threshold voltage distribution, erase-write times, and data retention time. However, some methods only focus on single characteristics such as NAND flash memory signal detection or life prediction, and do not fully consider multiple key characteristics of NAND flash memory. Some methods use neural network models, but the model structure is relatively simple and cannot fully capture the complex nonlinear relationships in NAND flash memory characteristics. These existing technologies still have some shortcomings, resulting in low accuracy of flash memory characteristic prediction. SUMMARY

[0005] The present application provides a flash memory characteristic prediction model training method, a flash memory characteristic prediction method and device to solve the problem of low accuracy of flash memory characteristic prediction.

[0006] In a first aspect, the present application provides a flash memory characteristic prediction model training method, comprising:

[0007] constructing graph structure data; the graph structure data comprises a plurality of nodes and edges between the nodes, the nodes are determined based on first target information of at least one type of flash memory, and the first target information comprises at least one of the following: manufacturing process information, electrical property information, and working condition information; and the edges represent the interaction relationship between the nodes.

[0008] training an initial flash memory characteristic prediction model based on the graph structure data to obtain a target flash memory characteristic prediction model.

[0009] According to the flash memory characteristic prediction model training method provided by the application, the graph structure data is constructed, which comprises:

[0010] obtaining the first target information of each type of flash memory;

[0011] performing characteristic testing on the first target information of each type of flash memory to obtain characteristic parameters and characteristic testing results in the characteristic testing process;

[0012] performing normalization processing on the first target information, the characteristic parameters, and the characteristic testing results respectively to obtain normalized first target information, normalized characteristic parameters, and normalized characteristic testing results;

[0013] constructing the graph structure data based on the normalized first target information, the normalized characteristic parameters, and the normalized characteristic testing results; the nodes are any one of the normalized first target information, the normalized characteristic parameters, and the normalized characteristic testing results.

[0014] According to the flash memory characteristic prediction model training method provided by the application, the initial flash memory characteristic prediction model comprises a graph attention network module and a full connection layer.

[0015] The initial flash memory characteristic prediction model is trained based on the graph structure data to obtain a target flash memory characteristic prediction model, which comprises:

[0016] Step A: inputting the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain target weights between any node in the graph attention network module output and other nodes in the neighborhood;

[0017] Step B: inputting each target weight into the full connection layer in the initial flash memory characteristic prediction model to obtain characteristic parameter prediction values of each type of flash memory output by the full connection layer;

[0018] Step C: determining the target flash memory characteristic prediction model based on the characteristic parameter predicted value of each of the flash memories and the characteristic parameter true value of each of the flash memories.

[0019] According to the flash memory characteristic prediction model training method provided by the application, the target flash memory characteristic prediction model is determined based on the characteristic parameter predicted value of each of the flash memories and the characteristic parameter true value of each of the flash memories, and the method comprises the following steps:

[0020] A loss function is calculated based on the characteristic parameter predicted value of each of the flash memories and the characteristic parameter true value of each of the flash memories.

[0021] In the case that the cumulative value of the loss function does not tend to be stable or the training times do not reach the maximum training times, the parameters in the initial flash memory characteristic prediction model are updated, and the above steps A to C are repeated until the cumulative value of the loss function tends to be stable or the training times reach the maximum training times.

[0022] The target flash memory characteristic prediction model is determined based on the parameters in the initial flash memory characteristic prediction model that are finally updated.

[0023] According to the flash memory characteristic prediction model training method provided by the application, the graph attention network module comprises at least one graph attention layer.

[0024] The graph structure data is input into the graph attention network module in the initial flash memory characteristic prediction model, and the target weight between any node in each of the nodes and other nodes in the neighborhood output by the graph attention network module is obtained.

[0025] The graph structure data is input into the first graph attention layer in the at least one graph attention layer included in the graph attention network module, and the first weight between any node in each of the nodes and other nodes in the neighborhood output by the first graph attention layer is obtained; the first weight is used to update the graph structure data.

[0026] The updated graph structure data is input into the second graph attention layer in the at least one graph attention layer, and the second weight between any node in each of the nodes and other nodes in the neighborhood output by the second graph attention layer is obtained; the first weight is used to update the updated graph structure data.

[0027] The above steps are repeatedly executed until the finally updated graph structure data is input into the last graph attention layer in the at least one graph attention layer, and the target weight between any node in each of the nodes and other nodes in the neighborhood output by the last graph attention layer is obtained.

[0028] In a second aspect, the present application provides a flash memory characteristic prediction method, comprising:

[0029] obtaining second target information of a target flash memory, the second target information comprising at least one of the following: manufacturing process information, electrical property information, and working condition information;

[0030] inputting the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model; the target flash memory characteristic prediction model being trained based on the flash memory characteristic prediction model training method of any one of the first aspect.

[0031] In a third aspect, the present application provides a flash memory characteristic prediction model training device, comprising:

[0032] a construction module configured to construct graph structure data; the graph structure data comprising a plurality of nodes and edges between the nodes, the nodes being determined based on first target information of at least one type of flash memory, the first target information comprising at least one of the following: manufacturing process information, electrical property information, and working condition information; the edges representing interaction relationships between the nodes;

[0033] a training module configured to train an initial flash memory characteristic prediction model based on the graph structure data to obtain a target flash memory characteristic prediction model.

[0034] In a fourth aspect, the present application provides a flash memory characteristic prediction device, comprising:

[0035] a first obtaining module configured to obtain second target information of a target flash memory, the second target information comprising at least one of the following: manufacturing process information, electrical property information, and working condition information;

[0036] a prediction module configured to input the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model; the target flash memory characteristic prediction model being trained based on the flash memory characteristic prediction model training device of any one of the third aspect.

[0037] In a fifth aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing the flash memory characteristic prediction model training method of any one of the first aspect or the flash memory characteristic prediction method of the second aspect when executing the computer program.

[0038] In a sixth aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the flash memory characteristic prediction model training method according to any one of the first aspect or the flash memory characteristic prediction method according to the second aspect.

[0039] In a seventh aspect, the present application also provides a computer program product comprising a computer program, which, when executed by a processor, implements the flash memory characteristic prediction model training method according to any one of the first aspect or the flash memory characteristic prediction method according to the second aspect.

[0040] The flash memory characteristic prediction model training method, the flash memory characteristic prediction method and the device provided by the present application can realize accurate prediction of flash memory characteristics and improve the accuracy of flash memory characteristic prediction by constructing graph structure data, training an initial flash memory characteristic prediction model based on the graph structure data, and obtaining a target flash memory characteristic prediction model. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0042] Figure 1 is one of the flowcharts of the flash memory characteristic prediction model training method provided by the present application.

[0043] Figure 2 is a schematic diagram of the graph structure data provided by the present application.

[0044] Figure 3 is a structural schematic diagram of the graph attention layer provided by the present application.

[0045] Figure 4 is another flowchart of the flash memory characteristic prediction model training method provided by the present application.

[0046] Figure 5 is a flowchart of the flash memory characteristic prediction method provided by the present application.

[0047] Figure 6 is a structural schematic diagram of a flash memory characteristic prediction model training device provided by the present application.

[0048] Figure 7 is a structural schematic diagram of a flash memory characteristic prediction device provided by the present application.

[0049] Figure 8 is a structural schematic diagram of an electronic device provided by the present application. DETAILED DESCRIPTION

[0050] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0051] The flash memory characteristic prediction model training method and the flash memory characteristic prediction method of the present application will be described below. Figures 1-5

[0052] Figure 1 is one of flowcharts of a flash memory characteristic prediction model training method provided by the present application, as shown in the figure, the method comprises the following steps 101-102. Figure 1

[0053] Step 101, constructing graph structure data; the graph structure data comprises a plurality of nodes and edges between the nodes, the nodes are determined based on first target information of at least one type of flash memory, the first target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information; the edges represent the interaction relationship between the nodes.

[0054] It should be noted that the flash memory characteristic prediction model training method provided by the present application can be applied to the scene of flash memory characteristic prediction, for example, NAND flash memory characteristic prediction, the execution subject of the method can be a flash memory characteristic prediction model training device, for example, an electronic device, or a control module in the flash memory characteristic prediction model training device for executing the flash memory characteristic prediction model training method.

[0055] ​​Specifically, the manufacturing process information is a structure of the flash memory or a material of the flash memory, the electrical property information is optimal read voltage axis information, and the working condition information is information of a time, a temperature, and an erase-write number of flash memory working. The graph structure data includes a plurality of nodes and edges between the nodes, wherein the nodes are determined based on first target information of at least one type of flash memory, and the first target information includes at least one of the following: the manufacturing process information, the electrical property information, and the working condition information; and the edges represent interaction relationships between the nodes.

[0056] In step 102, an initial flash memory characteristic prediction model is trained based on the graph structure data to obtain a target flash memory characteristic prediction model.

[0057] Specifically, the initial flash memory characteristic prediction model is trained based on the graph structure data to obtain the target flash memory characteristic prediction model.

[0058] The flash memory characteristic prediction model training method provided by the application trains an initial flash memory characteristic prediction model based on graph structure data, and obtains a target flash memory characteristic prediction model. The graph structure data is constructed by using multi-dimensional target information of various types of flash memory and interaction relationships therebetween, and the target flash memory characteristic prediction model trained therefrom can accurately predict flash memory characteristics and improve the accuracy of flash memory characteristic prediction.

[0059] Optionally, the specific implementation of step 101 includes:

[0060] The first target information of each type of flash memory is obtained, and the first target information of each type of flash memory is subjected to a characteristic test to obtain characteristic parameters and characteristic test results in the characteristic test process. The first target information, the characteristic parameters, and the characteristic test results are subjected to normalization processing to obtain normalized first target information, normalized characteristic parameters, and normalized characteristic test results. The graph structure data is constructed based on the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results. The nodes are any one of the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results.

[0061] Specifically, first target information of various types of flash memories (for example, NAND) is collected through data disclosed by various manufacturers, the first target information including at least one of manufacturing process information, electrical property information and working condition information; and then the first target information of various types of flash memories is subjected to a timing and optimal read voltage axis characteristic test, to obtain characteristic parameters and characteristic test results of the various types of flash memories in a characteristic test process in an actual working environment, the characteristic parameters including at least one of read / write times, temperature and data retention time, and the characteristic test results including at least one of write time and error times.

[0062] The first target information, the characteristic parameters and the characteristic test results are respectively preprocessed to clean up abnormal values and inconsistent data, so as to ensure the accuracy of subsequent analysis; then the first target information, the characteristic parameters and the characteristic test results are respectively normalized to obtain normalized first target information, normalized characteristic parameters and normalized characteristic test results, so as to improve the efficiency and accuracy of training of a flash memory characteristic prediction model; and then graph structure data is constructed based on the normalized first target information, the normalized characteristic parameters and the normalized characteristic test results, wherein a node is any one of the normalized first target information, the normalized characteristic parameters and the normalized characteristic test results.

[0063] Figure 2 is a schematic diagram of the graph structure data provided by the application, as shown in Figure 2 The nodes are working conditions, manufacturing processes, electrical properties, characteristic test results and characteristic parameters, and the working conditions, the manufacturing processes, the electrical properties, the characteristic test results and the characteristic parameters are connected through edges, the edges representing the interaction relationship between the working conditions, the manufacturing processes, the electrical properties, the characteristic test results and the characteristic parameters, but there is no edge between the characteristic parameters. Finally, feature embedding is performed on each node to convert the characteristic parameters into high-dimensional vector representation, so as to facilitate processing of an initial flash memory characteristic prediction model.

[0064] Optionally, the initial flash memory characteristic prediction model includes a graph attention network module and a full-link layer; and the specific implementation of the step 102 includes:

[0065] Step A: inputting the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain target weights between any node and other nodes in the neighborhood output by the graph attention network module.

[0066] It should be noted that in the training stage of the initial flash memory characteristic prediction model, the weights and biases of the initial flash memory characteristic prediction model are first initialized.

[0067] Specifically, the graph attention network module is used to determine the target weight between any node and other nodes in the neighborhood, and realize the weighted aggregation of neighbor nodes. By inputting the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model, the target weight between any node and other nodes in the neighborhood output by the graph attention network module can be obtained.

[0068] Step B: inputting each target weight into the full connection layer in the initial flash memory characteristic prediction model to obtain the characteristic parameter prediction value of each model flash memory output by the full connection layer.

[0069] Specifically, the full connection layer is used to output the characteristic parameter prediction value of each model flash memory. By inputting each target weight into the full connection layer in the initial flash memory characteristic prediction model, the characteristic parameter prediction value of each model flash memory output by the full connection layer can be obtained.

[0070] Step C: determining the target flash memory characteristic prediction model based on the characteristic parameter prediction value of each model flash memory and the characteristic parameter true value of each model flash memory.

[0071] Specifically, based on the characteristic parameter prediction value of each model flash memory and the characteristic parameter true value of each model flash memory, the target flash memory characteristic prediction model can be further determined.

[0072] Optionally, the graph attention network module comprises at least one graph attention layer; and the inputting of the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain the target weight between any node and other nodes in the neighborhood output by the graph attention network module comprises:

[0073] (1) inputting the graph structure data into the first graph attention layer of the at least one graph attention layer included in the graph attention network module to obtain the first weight between any node and other nodes in the neighborhood output by the first graph attention layer; and the first weight is used to update the graph structure data.

[0074] Specifically, the graph attention network module comprises at least one graph attention layer, and the graph attention layers are in series, that is, a plurality of graph attention layers are stacked to form the network body, each graph attention layer calculates the weight between nodes through an attention mechanism, and realizes weighted aggregation of neighbor nodes. The graph attention layer comprises steps of linear transformation, attention coefficient calculation and feature aggregation, and different weights are assigned to different nodes in the neighborhood; wherein, the graph attention layer adopts an attention mechanism, calculates the attention score between nodes by using an "attention head" structure, measures the similarity and importance between nodes by using a dot product attention and the like, updates the feature representation of the node according to the attention score, so that the node can capture richer context information, thereby assigning different weights to different nodes in the neighborhood.

[0075] Figure 3 is a structural schematic diagram of the graph attention layer provided by the application, as Figure 3 indicated, X0, X1, X2, X3 and X4 represent nodes, X0 is a center node, X1, X2, X3 and X4 are other nodes in the neighborhood of X0, a0,1 represents the weight between X1 and X0, a0,2 represents the weight between X2 and X0, a0,3 represents the weight between X3 and X0, and a0,4 represents the weight between X4 and X0; wherein, a0,1 is obtained by measuring the similarity and importance between nodes through a dot product attention and the like.

[0076] In the model training stage, the weights between nodes in the graph structure data need to be initialized. The graph structure data is input into the first graph attention layer in the at least one graph attention layer included in the graph attention network module, so that the first weight between any node in the nodes output by the first graph attention layer and other nodes in the neighborhood can be obtained; wherein, the first weight is used to update the graph structure data.

[0077] (2) the updated graph structure data is input into the second graph attention layer in the at least one graph attention layer, so that the second weight between any node in the nodes output by the second graph attention layer and other nodes in the neighborhood can be obtained; the first weight is used to update the updated graph structure data.

[0078] Specifically, after the first weight between any node in the nodes output by the first graph attention layer and other nodes in the neighborhood, the graph structure data is updated based on the first weight, so that the updated graph structure data is obtained. Then, the updated graph structure data is input into the second graph attention layer in the at least one graph attention layer, so that the second weight between any node in the nodes output by the second graph attention layer and other nodes in the neighborhood can be obtained; the first weight is used to update the updated graph structure data.

[0079] (3) repeating until the final updated graph structure data is input into a last graph attention layer of the at least one graph attention layer, to obtain the target weight between any node of each node output by the last graph attention layer and other nodes in the neighborhood.

[0080] Specifically, repeating until the final updated graph structure data is input into a last graph attention layer of the at least one graph attention layer, to obtain the target weight between any node of each node output by the last graph attention layer and other nodes in the neighborhood.

[0081] Optionally, the target flash memory characteristic prediction model is determined based on the characteristic parameter predicted value of each type of flash memory and the characteristic parameter true value of each type of flash memory, comprising:

[0082] (a) calculating a loss function based on the characteristic parameter predicted value of each type of flash memory and the characteristic parameter true value of each type of flash memory.

[0083] Specifically, the characteristic parameter true value of each type of flash memory is the characteristic parameter obtained in the characteristic test process, and the mean square error is used as the loss function to calculate the difference between the characteristic parameter predicted value and the characteristic parameter true value.

[0084] (b) in the case that the cumulative value of the loss function does not tend to be stable or the number of training times does not reach the maximum number of training times, updating the parameters in the initial flash memory characteristic prediction model, and repeating the above steps A to C until the cumulative value of the loss function tends to be stable or the number of training times reaches the maximum number of training times.

[0085] Specifically, it is judged whether the cumulative value of the loss function tends to be stable or the number of training times reaches the maximum number of training times. In the case that the cumulative value of the loss function does not tend to be stable or the number of training times does not reach the maximum number of training times, the gradient of the model parameters with respect to the loss function is used for back propagation, and the update of the network weight is guided, and finally the network weight is updated according to the gradient information obtained by back propagation using the Adam optimization algorithm. This process is iterative, each iteration contains a forward propagation and a back propagation, followed by the update of the weight, so as to update the parameters in the initial flash memory characteristic prediction model, and repeat the above steps A to C until the cumulative value of the loss function tends to be stable or the number of training times reaches the maximum number of training times. The maximum number of training times is pre-set, for example, the maximum number of training times is 100 times.

[0086] (c) determining the target flash memory characteristic prediction model based on the parameters in the initial flash memory characteristic prediction model updated finally.

[0087] Specifically, the parameters in the final updated initial flash memory characteristic prediction model are determined as the parameters of the target flash memory characteristic prediction model, so as to determine the target flash memory characteristic prediction model.

[0088] Figure 4 is a flowchart of the flash memory characteristic prediction model training method provided by the present application. Figure 4 As shown in the figure, the method comprises steps 401-411.

[0089] Step 401, data acquisition. Obtain first target information of each type of flash memory; the first target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information.

[0090] Step 402, characteristic test. Perform a characteristic test on the first target information of each type of flash memory to obtain characteristic parameters and characteristic test results during the characteristic test.

[0091] Step 403, data preprocessing. Perform cleaning and normalization processing on the first target information, the characteristic parameters and the characteristic test results, respectively, to obtain normalized first target information, normalized characteristic parameters and normalized characteristic test results.

[0092] Step 404, constructing graph structure data. Based on the normalized first target information, the normalized characteristic parameters and the normalized characteristic test results, construct graph structure data; the nodes are any one of the normalized first target information, the normalized characteristic parameters and the normalized characteristic test results.

[0093] Step 405, determining the first weight output by the first graph attention layer. Input the graph structure data into the first graph attention layer of at least one graph attention layer included in the graph attention network module to obtain the first weight between any node in the graph structure data and other nodes in the neighborhood; the first weight is used to update the graph structure data.

[0094] Step 406, determining the second weight output by the second graph attention layer. Input the updated graph structure data into the second graph attention layer of the at least one graph attention layer to obtain the second weight between any node in the graph structure data and other nodes in the neighborhood; the first weight is used to update the updated graph structure data.

[0095] Step 407, determining the target weight output by the last graph attention layer. Repeat until the final updated graph structure data is input into the last graph attention layer of the at least one graph attention layer to obtain the target weight between any node in the graph structure data and other nodes in the neighborhood.

[0096] Step 408, determine the characteristic parameter prediction value. The target weight is input into the full connection layer in the initial flash memory characteristic prediction model to obtain the characteristic parameter prediction value of each type of flash memory output by the full connection layer.

[0097] Step 409, calculate the loss function. Based on the characteristic parameter prediction value of each type of flash memory and the characteristic parameter true value of each type of flash memory, the loss function is calculated.

[0098] Step 410, update the parameters of the model and optimize. In the case that the cumulative value of the loss function does not tend to be stable or the training times do not reach the maximum training times, the parameters in the initial flash memory characteristic prediction model are updated, and the above steps 405 to 409 are repeated until the cumulative value of the loss function tends to be stable or the training times reach the maximum training times.

[0099] Step 411, determine the target flash memory characteristic prediction model. Based on the parameters in the finally updated initial flash memory characteristic prediction model, the target flash memory characteristic prediction model is determined.

[0100] The flash memory characteristic prediction model training method provided by the application fully considers the multi-dimensional characteristics of different types of flash memories and the complex relationship between them by constructing graph structure data, and then trains the flash memory characteristic prediction model through the graph structure data, so that the flash memory characteristic prediction model can adaptively learn the characteristic differences between different types of flash memories, and then the trained flash memory characteristic prediction model can still maintain high prediction performance when facing new types or new batches of flash memories. The improvement of this generalization ability helps to reduce the retraining and debugging cost of the model in different application scenarios, and can realize accurate prediction of the characteristics of the flash memory.

[0101] Figure 5 is the flowchart of the flash memory characteristic prediction method provided by the application, as Figure 5 shown, the method comprises steps 501-502.

[0102] Step 501, obtain the second target information of the target flash memory, and the second target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information.

[0103] Specifically, the manufacturing process information is the structure of the flash memory or the material of the flash memory, the electrical property information is the best read voltage axis information, and the working condition is the time, temperature and erase-write times of the flash memory working and the like.

[0104] Step 502, input the second target information into the target flash memory characteristic prediction model to obtain the characteristic parameters of the target flash memory output by the target flash memory characteristic prediction model.

[0105] The target flash memory characteristic prediction model is trained based on the flash memory characteristic prediction model training method in the foregoing embodiments.

[0106] Specifically, the second target information is input into the pre-trained target flash memory characteristic prediction model, so that a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model can be obtained. The characteristic parameter of the target flash memory includes at least one of the following: read-write times, temperature, and data retention time.

[0107] Optionally, the characteristic parameter of the target flash memory can also be output in the form of a visual chart or a report, so that a user can intuitively understand the current state (the characteristic parameter currently output) and the predicted trend of the target flash memory, which helps the user to better formulate a data storage and management strategy and optimize the allocation and use of storage resources.

[0108] Optionally, the accurate prediction of the characteristic of the target flash memory provides strong support for the optimization of storage technology. Manufacturers can improve the design and production process of the flash memory by using the predicted characteristic parameter, so as to improve the performance and reliability of the flash memory. Meanwhile, researchers can also conduct more in-depth research and development of storage technology based on the predicted characteristic parameter.

[0109] The flash memory characteristic prediction method provided by the application comprises the following steps: obtaining second target information of a target flash memory, wherein the second target information comprises at least one of the following: manufacturing process information, electrical property information, and working condition information; and inputting the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model. Since the multi-dimensional characteristics of different models of flash memories and the complex relationship therebetween are considered when the target flash memory characteristic prediction model is trained, the flash memory characteristic prediction model can adaptively learn the characteristic differences between different models of flash memories, so that the trained flash memory characteristic prediction model can realize accurate prediction of the characteristics of the flash memory and improve the read-write efficiency and stability of the SSD.

[0110] The flash memory characteristic prediction model training device provided by the application is described below. The flash memory characteristic prediction model training device described below can be correspondingly referred to the flash memory characteristic prediction model training method described above.

[0111] Figure 6 is a structural schematic diagram of the flash memory characteristic prediction model training device provided by the application, as Figure 6 shown, the flash memory characteristic prediction model training device 600 comprises a construction module 601 and a training module 602; wherein,

[0112] The constructing module 601 is configured to construct graph structure data, wherein the graph structure data comprises a plurality of nodes and edges between the nodes, the nodes are determined based on first target information of at least one type of flash memory, the first target information comprises at least one of manufacturing process information, electrical property information and working condition information, and the edges represent interaction relationships between the nodes.

[0113] The training module 602 is configured to train an initial flash memory characteristic prediction model based on the graph structure data to obtain a target flash memory characteristic prediction model.

[0114] The flash memory characteristic prediction model training device provided by the application trains an initial flash memory characteristic prediction model based on the graph structure data to obtain a target flash memory characteristic prediction model. The graph structure data is constructed by using multi-dimensional target information of various types of flash memory and interaction relationships therebetween, and the target flash memory characteristic prediction model obtained by training can realize accurate prediction of flash memory characteristics and improve the accuracy of flash memory characteristic prediction.

[0115] Optionally, the constructing module 601 is specifically configured to:

[0116] obtain the first target information of each type of flash memory;

[0117] perform characteristic testing on the first target information of each type of flash memory to obtain characteristic parameters and characteristic testing results in the characteristic testing process;

[0118] perform normalization processing on the first target information, the characteristic parameters and the characteristic testing results to obtain normalized first target information, normalized characteristic parameters and normalized characteristic testing results;

[0119] construct the graph structure data based on the normalized first target information, the normalized characteristic parameters and the normalized characteristic testing results, wherein the nodes are any one of the normalized first target information, the normalized characteristic parameters and the normalized characteristic testing results.

[0120] Optionally, the initial flash memory characteristic prediction model comprises a graph attention network module and a full-link layer, and the training module 602 is specifically configured to:

[0121] Step A: inputting the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain a target weight between any node in each of the nodes and other nodes in a neighborhood output by the graph attention network module;

[0122] Step B: inputting each of the target weights into the full-link layer in the initial flash memory characteristic prediction model to obtain a characteristic parameter prediction value of each of the model flash memories output by the full-link layer;

[0123] Step C: determining the target flash memory characteristic prediction model based on the characteristic parameter prediction value of each of the model flash memories and a characteristic parameter true value of each of the model flash memories.

[0124] Optionally, the training module 602 is further configured to:

[0125] calculating a loss function based on the characteristic parameter prediction value of each of the model flash memories and the characteristic parameter true value of each of the model flash memories;

[0126] in a case where the cumulative value of the loss function does not tend to be stable or the number of training times does not reach a maximum number of training times, updating parameters in the initial flash memory characteristic prediction model, and repeating the above steps A to C until the cumulative value of the loss function tends to be stable or the number of training times reaches the maximum number of training times;

[0127] determining the target flash memory characteristic prediction model based on the parameters in the initial flash memory characteristic prediction model that are finally updated.

[0128] Optionally, the graph attention network module comprises at least one graph attention layer; and the training module 602 is further configured to:

[0129] inputting the graph structure data into a first graph attention layer in the at least one graph attention layer included in the graph attention network module to obtain a first weight between any node in each of the nodes and other nodes in a neighborhood output by the first graph attention layer; the first weight is used to update the graph structure data;

[0130] inputting the updated graph structure data into a second graph attention layer in the at least one graph attention layer to obtain a second weight between any node in each of the nodes and other nodes in a neighborhood output by the second graph attention layer; the first weight is used to update the updated graph structure data;

[0131] repeatedly performing until finally updated graph structure data are inputted into a last graph attention layer in the at least one graph attention layer to obtain the target weight between any node in each of the nodes and other nodes in a neighborhood output by the last graph attention layer.

[0132] Figure 7 is a structural schematic diagram of a flash memory characteristic prediction device provided by the present application, as shown in Figure 7 The flash memory characteristic prediction device 700 comprises an acquisition module 701 and a prediction module 702; wherein,

[0133] The acquisition module 701 is configured to acquire second target information of a target flash memory, wherein the second target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information;

[0134] The prediction module 702 is configured to input the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model; and the target flash memory characteristic prediction model is obtained by training based on the flash memory characteristic prediction model training method shown in Figure 6

[0135] The flash memory characteristic prediction device provided by the present application acquires the second target information of the target flash memory, wherein the second target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information; and inputs the second target information into the target flash memory characteristic prediction model to obtain the characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model. Since the multi-dimensional characteristics of different models of flash memories and the complex relationship therebetween are considered when training the target flash memory characteristic prediction model, the flash memory characteristic prediction model can adaptively learn the characteristic differences between different models of flash memories, and thus the trained flash memory characteristic prediction model can realize accurate prediction of the characteristics of the flash memory, thereby improving the read-write efficiency and stability of the SSD.

[0136] Figure 8 is a structural schematic diagram of an electronic device provided by the present application, as shown in Figure 8 The electronic device can comprise a processor 810, a communications interface 820, a memory 830 and a communications bus 840, wherein the processor 810, the communications interface 820 and the memory 830 complete mutual communication through the communications bus 840. The processor 810 can invoke a logical instruction in the memory 830 to execute a flash memory characteristic prediction model training method, which comprises: constructing graph structure data; the graph structure data comprises a plurality of nodes and edges between the nodes, the nodes are determined based on first target information of at least one model of flash memory, the first target information comprises at least one of the following: manufacturing process information, electrical property information and working condition information; the edges represent the interaction relationship between the nodes; and based on the graph structure data, an initial flash memory characteristic prediction model is trained to obtain a target flash memory characteristic prediction model.​

[0137] The processor 810 can invoke the logic instructions in the memory 830 to execute a flash memory characteristic prediction method, which comprises: obtaining second target information of a target flash memory, the second target information comprising at least one of: manufacturing process information, electrical property information, and working condition information; inputting the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model; and the target flash memory characteristic prediction model is obtained by training based on the flash memory characteristic prediction model training method of any of the preceding embodiments.

[0138] In addition, the logic instructions in the memory 830 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0139] On the other hand, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor, so that the computer can execute the flash memory characteristic prediction model training method provided by the above-mentioned methods, which comprises: constructing graph structure data; the graph structure data comprises a plurality of nodes and edges between the nodes, the nodes are determined based on first target information of at least one type of flash memory, the first target information comprises at least one of: manufacturing process information, electrical property information, and working condition information; the edges represent the interaction relationship between the nodes; based on the graph structure data, an initial flash memory characteristic prediction model is trained to obtain a target flash memory characteristic prediction model.

[0140] The computer program is executed by the processor, and the computer can execute the flash memory characteristic prediction method provided by the above method, which comprises: obtaining second target information of a target flash memory, the second target information comprising at least one of the following: manufacturing process information, electrical property information and working condition information; inputting the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model; and the target flash memory characteristic prediction model is obtained by training the flash memory characteristic prediction model training method according to any one of the preceding embodiments.

[0141] In another aspect, the application also provides a non-transitory computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the flash memory characteristic prediction model training method provided by the above method, which comprises: constructing graph structure data; the graph structure data comprises a plurality of nodes and edges between the nodes, the nodes being determined based on first target information of at least one type of flash memory, the first target information comprising at least one of the following: manufacturing process information, electrical property information and working condition information; the edges represent the interaction relationship between the nodes; and based on the graph structure data, an initial flash memory characteristic prediction model is trained to obtain a target flash memory characteristic prediction model.

[0142] The computer program is executed by the processor, and the computer can execute the flash memory characteristic prediction method provided by the above method, which comprises: obtaining second target information of a target flash memory, the second target information comprising at least one of the following: manufacturing process information, electrical property information and working condition information; inputting the second target information into a target flash memory characteristic prediction model to obtain a characteristic parameter of the target flash memory output by the target flash memory characteristic prediction model; and the target flash memory characteristic prediction model is obtained by training the flash memory characteristic prediction model training method according to any one of the preceding embodiments.

[0143] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment. Those skilled in the art can understand and implement without creative labor.

[0144] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0145] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for training a flash memory characteristic prediction model, characterized in that, include: Constructing graph structure data; the graph structure data includes multiple nodes and edges between the nodes, the nodes are determined based on first target information of at least one type of flash memory, the first target information including at least one of the following: manufacturing process information, electrical attribute information, and operating condition information; the edges represent the interaction relationships between the nodes; Based on the graph structure data, the initial flash memory characteristic prediction model is trained to obtain the target flash memory characteristic prediction model; The constructed graph structure data includes: Obtain the first target information for each of the aforementioned flash memory models; The first target information of each of the aforementioned flash memory models is subjected to characteristic testing to obtain characteristic parameters and characteristic test results during the characteristic testing process; The first target information, the characteristic parameters, and the characteristic test results are normalized respectively to obtain the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results. Based on the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results, the graph structure data is constructed; the node is any one of the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results. The initial flash memory characteristic prediction model includes a graph attention network module and a fully connected layer; The step of training the initial flash memory characteristic prediction model based on the graph structure data to obtain the target flash memory characteristic prediction model includes: Step A: Input the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain the target weight between any node and other nodes in the neighborhood of each node output by the graph attention network module; Step B: Input each of the target weights into the fully linked layer in the initial flash memory characteristic prediction model to obtain the predicted characteristic parameters of each type of flash memory output by the fully linked layer; Step C: Based on the predicted characteristic parameters of each type of flash memory and the actual characteristic parameters of each type of flash memory, determine the target flash memory characteristic prediction model.

2. The flash memory characteristic prediction model training method according to claim 1, characterized in that, The step of determining the target flash memory characteristic prediction model based on the predicted characteristic parameters of each of the aforementioned flash memory models and the actual characteristic parameters of each of the aforementioned flash memory models includes: Based on the predicted values ​​of the characteristic parameters of each type of flash memory and the actual values ​​of the characteristic parameters of each type of flash memory, a loss function is calculated; If the cumulative value of the loss function does not stabilize or the number of training iterations has not reached the maximum number of training iterations, update the parameters in the initial flash memory feature prediction model and repeat steps A to C above until the cumulative value of the loss function stabilizes or the number of training iterations reaches the maximum number of training iterations. The target flash memory characteristic prediction model is determined based on the parameters in the final updated initial flash memory characteristic prediction model.

3. The flash memory characteristic prediction model training method according to claim 1, characterized in that, The graph attention network module includes at least one graph attention layer; The step of inputting the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain the target weight between any node and other nodes in the neighborhood of each node output by the graph attention network module includes: The graph structure data is input to the first graph attention layer in the at least one graph attention layer included in the graph attention network module to obtain a first weight between any node in each of the nodes output by the first graph attention layer and other nodes in the neighborhood; the first weight is used to update the graph structure data; The updated graph structure data is input into the second graph attention layer in the at least one graph attention layer to obtain a second weight between any node in each of the nodes output by the second graph attention layer and other nodes in the neighborhood; the first weight is used to update the updated graph structure data. Repeat the process until the final updated graph structure data is input into the last graph attention layer of the at least one graph attention layer, to obtain the target weight between any node in each of the nodes output by the last graph attention layer and other nodes in the neighborhood.

4. A method for predicting flash memory characteristics, characterized in that, include: Acquire second target information of the target flash memory, the second target information including at least one of the following: manufacturing process information, electrical attribute information, and operating condition information; The second target information is input into the target flash memory characteristic prediction model to obtain the characteristic parameters of the target flash memory output by the target flash memory characteristic prediction model; the target flash memory characteristic prediction model is trained based on the flash memory characteristic prediction model training method according to any one of claims 1 to 3.

5. A flash memory characteristic prediction model training device, characterized in that, include: A construction module is used to construct graph structure data; the graph structure data includes multiple nodes and edges between the nodes, the nodes are determined based on first target information of at least one type of flash memory, the first target information includes at least one of the following: manufacturing process information, electrical attribute information, and operating condition information; the edges represent the interaction relationships between the nodes. The training module is used to train the initial flash memory characteristic prediction model based on the graph structure data to obtain the target flash memory characteristic prediction model. The building module is specifically used for: Obtain the first target information for each of the aforementioned flash memory models; The first target information of each of the aforementioned flash memory models is subjected to characteristic testing to obtain characteristic parameters and characteristic test results during the characteristic testing process; The first target information, the characteristic parameters, and the characteristic test results are normalized respectively to obtain the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results. Based on the normalized first target information, the normalized characteristic parameters, and the normalized characteristic test results, the graph structure data is constructed. The node is any one of the normalized first target information, the normalized characteristic parameter, and the normalized characteristic test result; The initial flash memory characteristic prediction model includes a graph attention network module and a fully connected layer; The training module is specifically used for: Step A: Input the graph structure data into the graph attention network module in the initial flash memory characteristic prediction model to obtain the target weight between any node and other nodes in the neighborhood of each node output by the graph attention network module; Step B: Input each of the target weights into the fully linked layer in the initial flash memory characteristic prediction model to obtain the predicted characteristic parameters of each type of flash memory output by the fully linked layer; Step C: Based on the predicted characteristic parameters of each type of flash memory and the actual characteristic parameters of each type of flash memory, determine the target flash memory characteristic prediction model.

6. A flash memory characteristic prediction device, characterized in that, include: The acquisition module is used to acquire second target information of the target flash memory, the second target information including at least one of the following: manufacturing process information, electrical attribute information, and operating condition information; The prediction module is used to input the second target information into the target flash memory characteristic prediction model to obtain the characteristic parameters of the target flash memory output by the target flash memory characteristic prediction model; the target flash memory characteristic prediction model is trained based on the flash memory characteristic prediction model training method according to any one of claims 1 to 3.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the flash memory characteristic prediction model training method as described in any one of claims 1 to 3, or implements the flash memory characteristic prediction method as described in claim 4.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the flash memory characteristic prediction model training method as described in any one of claims 1 to 3, or implements the flash memory characteristic prediction method as described in claim 4.

Images