Adaptive operator parallel partitioning method for heterogeneous embedded chips in intelligent Internet of Things

Abstract

The present invention discloses an adaptive operator parallel partitioning method for heterogeneous embedded chips of intelligent Internet of Things. First, energy consumption is predicted based on an energy consumption prediction model of a random forest regression algorithm. For a variety of hardware environments and task characteristics, an energy consumption data set is constructed and feature selection is performed to achieve real-time monitoring of device status and task requirements and accurate energy consumption prediction. Secondly, an operator adaptive partitioning method based on a dynamic programming algorithm is used for efficient parallel reasoning of deep neural networks on multiple heterogeneous processors. On the basis of energy consumption prediction, the dependencies between operators, data communication overhead across processors, and dynamic resource conditions of processors are comprehensively considered to ensure that the system achieves optimal energy efficiency while meeting performance requirements. Through the above methods, the present invention significantly improves the energy efficiency and response speed of deep neural network reasoning on mobile devices, and solves the problem that static partitioning methods in the prior art are difficult to adapt to dynamic environments.

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet of Things, and in particular relates to an adaptive operator parallel partitioning method for heterogeneous embedded chips of intelligent Internet of Things. Background Art

[0002] With the advancement of science and technology and the rapid development of society and the economy, automated and intelligent technologies have become an inevitable trend in modern society. Smart IoT devices, particularly mobile devices like smartphones, have become widely adopted worldwide and are gradually permeating every aspect of daily life. While these devices offer a wealth of functions and applications, they also place higher demands on their internal computing resources and energy efficiency management.

[0003] The deployment and application of deep neural networks (DNNs) on mobile devices has become a hot topic of research. DNNs have demonstrated powerful capabilities in a variety of fields, including image recognition, speech recognition, natural language processing, and augmented reality. However, the high computational complexity and resource requirements of DNNs pose significant challenges to energy efficiency management in battery-powered mobile devices.

[0004] Currently, when implementing energy-efficient parallel DNN inference, existing static partitioning schemes struggle to adapt to the changing loads and processor conditions of mobile devices, resulting in decreased performance and energy efficiency. To overcome these challenges, existing research has proposed several solutions. For example, model compression, dynamic voltage and frequency scaling (DVFS), and memory access optimization can improve energy efficiency to a certain extent. However, these approaches often focus on a single processor or static environment and cannot fully address the complex interactions between heterogeneous processors and the dynamically changing hardware environment. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, the present invention provides an adaptive operator parallel partitioning method for heterogeneous embedded chips of intelligent Internet of Things. First, an energy consumption prediction model based on the random forest regression algorithm is used for energy consumption prediction in a dynamic environment of mobile devices; for a variety of hardware environments and task characteristics, an energy consumption data set is constructed and feature selection is performed. Through offline modeling and online adjustment, real-time monitoring of device status and task requirements and accurate energy consumption prediction are achieved. Secondly, an operator adaptive partitioning method based on a dynamic programming algorithm is used for efficient parallel reasoning of deep neural networks on multiple heterogeneous processors. On the basis of energy consumption prediction, the dependency between operators, data communication overhead across processors, and dynamic resource conditions of processors are comprehensively considered to optimize the operator partitioning strategy in real time to ensure that the system achieves optimal energy efficiency while meeting performance requirements. Through the above methods, the present invention significantly improves the energy efficiency and response speed of deep neural network reasoning on mobile devices, and solves the problem that static partitioning methods in the prior art are difficult to adapt to dynamic environments.

[0006] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0007] Step 1: Construction of dynamic environment energy consumption dataset of mobile devices;

[0008] Collect energy consumption data in various hardware environments for different mobile devices and heterogeneous processors;

[0009] Record the collected data according to the timestamp to ensure the continuity and integrity of the data;

[0010] Preprocess the collected data;

[0011] Normalize the data;

[0012] Finally, the processed data is stored in the specified format to construct an energy consumption dataset for mobile devices in a dynamic environment;

[0013] Step 2: Design of feature selection method for dynamic environment energy consumption prediction of mobile devices;

[0014] Use random forest regression algorithm for feature selection;

[0015] Assume that the energy consumption data set is D = {(X i ,y i )}, where X i is the eigenvector, y i For the corresponding energy consumption value, the goal of feature selection is to find a set of features F that minimizes the error of the energy consumption prediction model:

[0016]

[0017] in, The energy consumption value predicted by the model;

[0018] Step 3: Energy consumption prediction model training in dynamic environment;

[0019] The random forest regression model is used to predict the energy consumption of mobile devices in a dynamic environment. The energy consumption prediction error is optimized during the training process. The prediction error calculation formula is as follows:

[0020]

[0021] Among them, |D test | is the test data set, y true,i is the actual energy consumption value, y pred,i The energy consumption value predicted by the model;

[0022] By continuously adjusting the model parameters to minimize the above errors, the optimal energy consumption prediction model is finally obtained;

[0023] Step 4: Operator adaptive partitioning based on dynamic programming;

[0024] Introducing energy consumption feedback to optimize operator allocation by minimizing total energy consumption;

[0025] Considering the dependencies between operators and the dynamic resource conditions of the processor, the dynamic planning table is updated to calculate the execution time and energy consumption of each operator on different processors;

[0026] Use the minimum energy consumption value in the dynamic programming table to determine the optimal operator partitioning strategy and continuously adjust it to adapt to the dynamic environment;

[0027] DP[i][j]=min{DP[i-1][k]+E ij (t)}

[0028] Among them, DP[i][j] represents the minimum energy consumption of allocating the first i operators to the first j processors, E ij (t) represents the operator v i On the processor P j energy consumption;

[0029] Step 5: Operator scheduling based on adaptive adjustment algorithm;

[0030] Adaptive adjustment algorithms are used to adjust operator partitioning strategies based on current resource status and historical status to achieve energy efficiency optimization.

[0031] ΔR=∥R(t)-R k ∥

[0032] Among them, R(t) is the current resource state, R k Historical resource status, ΔR is the deviation between the current status and the historical status;

[0033] By calculating the deviation between the current state and the historical state, it is determined whether the operator partitioning strategy needs to be adjusted. When the deviation is less than the preset threshold, the current operator partitioning strategy remains unchanged; otherwise, the operator partitioning strategy is adjusted according to the current resource state.

[0034] Preferably, the energy consumption data set includes a feature vector and a corresponding energy consumption value.

[0035] Preferably, the feature vector includes the usage status and task characteristics of the processor.

[0036] Preferably, the usage status of the processor includes the usage rate and frequency of the CPU, GPU, and DSP.

[0037] Preferably, the task characteristics include operator type and operator size.

[0038] The beneficial effects of the present invention are as follows:

[0039] (1) The present invention proposes an adaptive operator parallel partitioning method for heterogeneous embedded chips in intelligent Internet of Things. By combining the random forest regression algorithm and the dynamic programming algorithm, it can accurately predict energy consumption in a dynamic environment and adaptively adjust the operator partitioning strategy, significantly improving the energy efficiency of deep neural network reasoning.

[0040] (2) This invention uses an operator scheduling method based on dynamic programming and adaptive adjustment to achieve efficient parallel operation of operators on multiple heterogeneous processors, optimize the performance of deep neural network reasoning, and reduce execution time and energy consumption.

[0041] (3) The present invention monitors the device status and task requirements in real time, utilizes energy consumption feedback and state deviation calculation, and adaptively adjusts the operator partitioning strategy, thereby being able to dynamically adapt to the changing hardware environment and maintain efficient operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is the overall framework diagram of the adaptive operator parallel partitioning method for heterogeneous embedded chips in the intelligent Internet of Things in the example of the present invention.

[0043] Figure 2 This is a diagram of the energy consumption prediction algorithm based on random forest regression in the present invention.

[0044] Figure 3 This is the dynamic programming optimization algorithm diagram of the present invention.

[0045] Figure 4 This is a diagram of the adaptive adjustment algorithm of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described below with reference to the accompanying drawings and examples.

[0047] This paper provides an adaptive operator parallel partitioning method for heterogeneous embedded chips in intelligent IoT, enabling energy-efficient deep neural network inference systems. In real-world mobile device scenarios, an intelligent scheduling model based on random forest regression and dynamic programming algorithms is proposed.

[0048] like Figures 1 to 4 The technical solution adopted by the present invention to solve the technical problem includes the following steps:

[0049] Step 1: Construction of dynamic environment energy consumption dataset of mobile devices;

[0050] For different mobile devices and heterogeneous processors, energy consumption data is collected in various hardware environments to build an energy consumption dataset for mobile devices in a dynamic environment. The collected data includes but is not limited to the processor usage status (such as the usage rate and frequency of the CPU, GPU, and DSP), task characteristics (such as operator type and operator size), etc.

[0051] Step 2: Design of feature selection method for dynamic environment energy consumption prediction of mobile devices;

[0052] Use random forest regression algorithm for feature selection. Assume that the energy consumption dataset is D={(X i ,y i )}, where X i is the eigenvector, y i The goal of feature selection is to find a set of features F that minimizes the error of the energy consumption prediction model:

[0053]

[0054] in, The energy consumption value predicted by the model;

[0055] Step 3: Energy consumption prediction model training in dynamic environment:

[0056] The random forest regression model is used to predict the energy consumption of mobile devices in a dynamic environment. The main goal during the training process is to optimize the energy consumption prediction error. The prediction error calculation formula is as follows:

[0057]

[0058] Among them, |D test | is the test data set, y true,i is the actual energy consumption value, y pred,i The energy consumption value predicted by the model is obtained by continuously adjusting the model parameters to minimize the above error and finally obtain the optimal energy consumption prediction model.

[0059] Step 4: Operator adaptive partitioning based on dynamic programming;

[0060] Introducing energy consumption feedback to optimize operator allocation by minimizing total energy consumption;

[0061] Considering the dependencies between operators and the dynamic resource conditions of the processor, the dynamic planning table is updated to calculate the execution time and energy consumption of each operator on different processors;

[0062] Use the minimum energy consumption value in the dynamic programming table to determine the optimal operator partitioning strategy and continuously adjust it to adapt to the dynamic environment;

[0063] DP[i][j]=min{DP[i-1][k]+Eij (t)}

[0064] Where DP[i][j] represents the minimum energy consumption of allocating the first i operators to the first j processors. ij (t) represents the operator v i On the processor P j energy consumption.

[0065] Step 5: Operator scheduling based on adaptive adjustment algorithm;

[0066] Adaptive adjustment algorithms are used to adjust operator partitioning strategies based on current resource status and historical status to achieve energy efficiency optimization.

[0067] ΔR=∥R(t)-R k ∥

[0068] Among them, R(t) is the current resource state, R k Historical resource status, ΔR is the deviation between the current status and the historical status.

[0069] Specifically, the operator partitioning strategy is determined by calculating the deviation between the current state and the historical state. If the deviation is less than a preset threshold, the current operator partitioning strategy remains unchanged. Otherwise, the operator partitioning strategy is adjusted based on the current resource state.

[0070] Example:

[0071] The present invention proposes an adaptive operator parallel partitioning method for heterogeneous embedded chips in intelligent Internet of Things, aiming to improve the inference energy efficiency and performance of deep neural networks (DNNs) on mobile devices. Through energy consumption prediction and operator adaptive partitioning in a dynamic environment, efficient parallel inference of DNNs on multiple heterogeneous processors is achieved. During the implementation process, an energy consumption prediction model is first constructed based on the random forest regression algorithm for energy consumption prediction in the dynamic environment of mobile devices. By collecting energy consumption data under various hardware environments, an energy consumption dataset is constructed, and feature selection and model training are performed. Then, the operators of DNN are adaptively divided and scheduled based on the dynamic programming algorithm, and the operator partitioning strategy is optimized in real time by comprehensively considering the dependencies between operators, the data communication overhead across processors, and the dynamic resource conditions of the processors.

[0072] To build the model dataset, we collected energy consumption data from various hardware environments for different mobile devices and heterogeneous processors to construct an energy consumption dataset for mobile devices in dynamic environments. The collected data includes, but is not limited to, processor usage status (such as CPU, GPU, and DSP utilization and frequency) and task characteristics (such as operator type and size).

[0073] For model feature selection and training, a random forest regression algorithm is used to extract key features relevant to energy consumption prediction, ensuring the efficiency of model input features. Combining offline modeling with online adjustments, the random forest regression model provides real-time monitoring of device status and task requirements, as well as energy consumption prediction, ensuring accurate and timely energy consumption predictions.

[0074] For adaptive operator partitioning and scheduling, a dynamic programming algorithm is used to adaptively partition and schedule DNN operators. This algorithm comprehensively considers inter-operator dependencies, cross-processor data communication overhead, and dynamic processor resource availability, optimizing the operator partitioning strategy in real time. An adaptive adjustment algorithm is used to adjust the operator partitioning strategy based on current and historical resource status to achieve energy efficiency optimization.

[0075] Through the above method, the present invention improves the energy efficiency and response speed of deep neural network reasoning on mobile devices.

[0076] To achieve the above tasks, the present invention adopts the following technical solutions:

[0077] An adaptive operator parallel partitioning method for heterogeneous embedded chips in intelligent Internet of Things includes the following steps:

[0078] Step 1: Constructing a dynamic energy consumption dataset for mobile devices: Collect processor usage and energy consumption data under different hardware environments and task loads. Task loads can include various deep neural network inference tasks, such as image recognition and speech recognition. The collected data is recorded with timestamps to ensure data continuity and integrity. The recorded data should include processor usage, task characteristics, and corresponding energy consumption values. Preprocess the collected data, including noise removal and missing value filling, to ensure data quality. Normalize the data to make it suitable for subsequent feature selection and model training. Finally, store the processed data in a specific format to construct an energy consumption dataset. The dataset should include feature vectors and corresponding energy consumption values. The feature vectors include processor usage (such as CPU, GPU, and DSP utilization and frequency), task characteristics (such as operator type and size), and corresponding energy consumption values.

[0079] Step 2: Design of feature selection method for dynamic environment energy consumption prediction of mobile devices;

[0080] Use random forest regression algorithm for feature selection. Assume that the energy consumption dataset is D={(X i ,y i )}, where X i is the eigenvector, y i The goal of feature selection is to find a set of features F that minimizes the error of the energy consumption prediction model:

[0081]

[0082] in, The energy consumption value predicted by the model;

[0083] During feature selection, data preprocessing is first required, including data cleaning and normalization to eliminate noise and outliers. The importance of each feature is then assessed using the random forest regression algorithm. Random forest constructs multiple decision trees and evaluates the splitting effect of each feature within each tree, ultimately generating a feature importance score. Specifically, feature importance is determined by calculating the average reduced error for each feature at its split points across all trees. Based on the feature importance score, the features most influential for energy consumption prediction are selected. A threshold is set, and all features with an importance score greater than the threshold are selected. Selected features include processor utilization, frequency, temperature, task partitioning ratio, number of channels, and number of cores. These features comprehensively reflect the processor's status and task characteristics, thereby improving the accuracy and robustness of the energy consumption prediction model.

[0084] Step 3: Energy consumption prediction model training in dynamic environment:

[0085] A random forest regression model is used to predict energy consumption in a dynamic environment for mobile devices. The training process primarily optimizes the energy prediction error. First, a training dataset and a test dataset, which have undergone feature selection, are prepared. The training dataset is used to train the random forest regression model, while the test dataset is used to evaluate the model's prediction performance. During training, the random forest regression model is fitted to the training data to construct an energy prediction model. The goal of model training is to optimize the model parameters by minimizing the energy prediction error.

[0086] The calculation formula of the prediction error is as follows: The calculation formula of the prediction error is as follows:

[0087]

[0088] Among them, |D test | is the test data set, y true,i is the actual energy consumption value, y pred,i The energy consumption value predicted by the model is obtained by continuously adjusting the model parameters to minimize the above error and finally obtain the optimal energy consumption prediction model.

[0089] First, the model parameters are initialized. The parameters of the random forest regression model, such as the number of decision trees and the maximum depth, are selected to ensure that the model has sufficient complexity to capture patterns in the data. Then, the random forest regression model is trained using the training dataset. The model constructs multiple decision trees and evaluates the splitting effect of each feature in each tree, ultimately generating a prediction of energy consumption. Next, the model's predictive performance is evaluated using the test dataset. The prediction error of the model on the test dataset is calculated, and the model parameters are adjusted based on the error. Model parameters are adjusted based on the magnitude of the prediction error. If the error is large, the model parameters are reselected and the training and evaluation process is repeated until the error is minimized.

[0090] Step 4: Operator adaptive partitioning based on dynamic programming;

[0091] Introducing energy consumption feedback to optimize operator allocation by minimizing total energy consumption;

[0092] Considering the dependencies between operators and the dynamic resource conditions of the processor, the dynamic planning table is updated to calculate the execution time and energy consumption of each operator on different processors;

[0093] Use the minimum energy consumption value in the dynamic programming table to determine the optimal operator partitioning strategy and continuously adjust it to adapt to the dynamic environment;

[0094] DP[i][j]=min{DP[i-1][k]+E ij (t)}

[0095] Where DP[i][j] represents the minimum energy consumption of allocating the first i operators to the first j processors. ij (t) represents the operator v i On the processor P j energy consumption.

[0096] During the calculation process, the allocation strategy is first initialized, and the execution time and energy consumption of each operator on the processor are calculated. Next, the optimal allocation strategy for each operator is updated using a dynamic programming table. By gradually selecting the optimal operator partitioning scheme, the system ensures that the performance requirements are met while achieving optimal energy efficiency. Specifically, all operators and processors are first traversed, the execution time and energy consumption of the operator on each processor are calculated, and the results are stored in the dynamic programming table. The optimal operator partitioning strategy is then found by gradually updating the values ​​in the table. At each step, the allocation scheme with the lowest energy consumption is selected to ensure optimal system energy efficiency.

[0097] Step 5: Operator scheduling based on adaptive adjustment algorithm;

[0098] Adaptive adjustment algorithms are used to adjust operator partitioning strategies based on current resource status and historical status to achieve energy efficiency optimization.

[0099] Specifically, by calculating the deviation between the current state and the historical state, we determine whether the operator partitioning strategy needs to be adjusted. If the deviation is less than the preset threshold, the current operator partitioning strategy remains unchanged; otherwise, the operator partitioning strategy is adjusted based on the current resource state.

[0100] ΔR=∥R(t)-R k ∥

[0101] Among them, R(t) is the current resource state, R k Historical resource status, ΔR is the deviation between the current status and the historical status.

[0102] The workflow of operator scheduling based on the adaptive adjustment algorithm begins with resource status monitoring. The current resource status includes dynamic parameters such as battery charge, processor load, and processor frequency. The adaptive adjustment algorithm utilizes this real-time data to evaluate the suitability of the current operator partitioning strategy. Before performing adaptive adjustments, the module compares the current resource status with previously saved execution plans and corresponding states. These historical execution plans and states contain operator partitioning strategies used in the past under different resource conditions and their performance. By calculating the deviation between the current resource status and the most similar historical state, the system determines the degree of similarity between the current state and the historical state. The deviation is calculated based on the difference between the current resource state and the historical state. If the deviation is small, that is, less than a preset threshold, it indicates that the current state is similar to the historical state, and the current operator partitioning strategy is still applicable, so the current strategy remains unchanged. This approach avoids frequent strategy adjustments and reduces system overhead. If the deviation is large, it indicates that the current resource state has changed significantly, and the existing operator partitioning strategy may no longer be applicable. In this case, the adaptive adjustment module needs to dynamically adjust the operator partitioning strategy based on the current resource state. By recalculating the execution time and energy consumption of each operator on each processor and using the adjustment function to update the operator partition ratio, we ensure that the new strategy can be adapted to the current resource status.

Claims

1. An adaptive operator parallel partitioning method for heterogeneous embedded chips in intelligent Internet of Things, characterized by: The steps include: Step 1: Construction of dynamic environment energy consumption dataset of mobile devices; Collect energy consumption data in various hardware environments for different mobile devices and heterogeneous processors; Record the collected data according to the timestamp to ensure the continuity and integrity of the data; Preprocess the collected data; Normalize the data; Finally, the processed data is stored in the specified format to construct an energy consumption dataset for mobile devices in a dynamic environment; Step 2: Design of feature selection method for dynamic environment energy consumption prediction of mobile devices; Use random forest regression algorithm for feature selection; Assume that the energy consumption data set is D = {(X i ,y i )}, where X i is the eigenvector, y i For the corresponding energy consumption value, the goal of feature selection is to find a set of features F that minimizes the error of the energy consumption prediction model: in, The energy consumption value predicted by the model; Step 3: Energy consumption prediction model training in dynamic environment; The random forest regression model is used to predict the energy consumption of mobile devices in a dynamic environment. The energy consumption prediction error is optimized during the training process. The prediction error calculation formula is as follows: Among them, |D test | is the test data set, y true,i is the actual energy consumption value, y pred,i The energy consumption value predicted by the model; By continuously adjusting the model parameters to minimize the above errors, the optimal energy consumption prediction model is finally obtained; Step 4: Operator adaptive partitioning based on dynamic programming; Introducing energy consumption feedback to optimize operator allocation by minimizing total energy consumption; Considering the dependencies between operators and the dynamic resource conditions of the processor, the dynamic planning table is updated to calculate the execution time and energy consumption of each operator on different processors; Use the minimum energy consumption value in the dynamic programming table to determine the optimal operator partitioning strategy and continuously adjust it to adapt to the dynamic environment; DP[i][j]=min{DP[i-1][k]+E ij (t)} Among them, DP[i][j] represents the minimum energy consumption of allocating the first i operators to the first j processors, E ij (t) represents the operator v i On the processor P j energy consumption; Step 5: Operator scheduling based on adaptive adjustment algorithm; Adaptive adjustment algorithms are used to adjust operator partitioning strategies based on current resource status and historical status to achieve energy efficiency optimization. ΔR=∥R(t)-R k ∥ Among them, R(t) is the current resource state, R k Historical resource status, ΔR is the deviation between the current status and the historical status; By calculating the deviation between the current state and the historical state, it is determined whether the operator partitioning strategy needs to be adjusted. When the deviation is less than the preset threshold, the current operator partitioning strategy remains unchanged; otherwise, the operator partitioning strategy is adjusted according to the current resource state.

2. The adaptive operator parallel partitioning method for heterogeneous embedded chips of intelligent Internet of Things according to claim 1 is characterized in that: The energy consumption data set includes a feature vector and a corresponding energy consumption value.

3. The adaptive operator parallel partitioning method for heterogeneous embedded chips of intelligent Internet of Things according to claim 1 is characterized in that: The feature vector includes the usage status and task features of the processor.

4. The adaptive operator parallel partitioning method for heterogeneous embedded chips of intelligent Internet of Things according to claim 1 is characterized in that: The processor usage status includes the usage rate and frequency of the CPU, GPU, and DSP.

5. The method for adaptive operator parallel partitioning of heterogeneous embedded chips for intelligent Internet of Things according to claim 1, characterized in that: Task characteristics include operator type and operator size.

Images