117 results about "Uniform quantization" patented technology

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as medical health and financial science and technology, and discloses a model quantitative reasoning acceleration method, device, equipment and medium. The method comprises the steps that an input text is divided into a plurality of processing blocks, importance scoring is conducted on the non-first processing block, and calculation precision formats are distributed according to scoring results; determining a unified quantization configuration of each processing block; dividing the network modules into configuration sharing groups, and sharing quantitative configurations of corresponding processing blocks in the groups; and executing block-level quantization inference according to the unified quantization configuration, and generating a model inference result. The quantitative configuration of each processing block is uniformly determined on the basis of the token importance score, and the configuration is multiplexed in the network module group, so that block-level precision distribution and parallel quantitative reasoning are realized, the video memory overhead and the configuration time overhead are greatly reduced while the reasoning precision is guaranteed, and the reasoning efficiency is improved. And the execution efficiency and the video memory utilization rate in the long text reasoning task are effectively improved.

The invention relates to the technical field of high-performance computing resource management, and discloses a computing resource allocation method for a distributed supercomputing center. The method comprises the following steps: on the basis of obtaining real-time computing task and supercomputing center resource data and uniformly quantifying, integrally predicting resource requirements of future tasks; constructing a mixed integer linear programming model with the minimization of the total operation cost as a single target, wherein the total operation cost is the sum of the energy cost, the carbon emission cost, the data transmission cost and the SLA default penalty cost; solving the model by taking the time-varying electricity price, the green energy ratio, the resource capacity and the network parameters of each center as constraint conditions to generate an optimal resource allocation scheme; and then, by dynamically monitoring the resource state and the task progress, the model is triggered to resolve when the resource utilization rate is detected to be unbalanced or default risks, so that self-adaptive adjustment is realized. According to the invention, global collaborative resource allocation across super computing centers is realized, and operation economy, environmental sustainability and service reliability are considered.

The invention discloses a high-precision quantization method for non-uniformly distributed data, and aims to solve the problems that reconstruction errors are remarkably increased and codebook resources are wasted when a traditional uniform quantization algorithm processes complex data distribution. According to the scheme, firstly, a dynamic programming algorithm is used for carrying out local homogenization processing on data, and optimal segmentation is achieved by minimizing the sum of squares of a weighting range; and then introducing a particle swarm optimization algorithm to carry out iterative tuning on the end points of the quantization interval, and constructing an efficient codebook by taking the minimization of a mean square error (MSE) as a target. Experimental results show that the quantization errors of the method on a 128-dimensional data set and a 420-dimensional data set are respectively reduced by 93.52% and 98.05% compared with those of a traditional method, meanwhile, hardware compatibility is kept, and the method is suitable for scenes such as edge computing nodes, AI chips and the like which have strict requirements on compression efficiency and real-time performance.

The invention provides a bandwidth distortion self-adaptive multi-layer scalable feature compressor and a method and system thereof, and relates to the technical field of machine learning. According to the invention, through a code rate distortion joint controller, bit width distribution and fragment size of each layer of a layering number are dynamically decided according to a real-time network state training stage and a precision demand; the input features are mapped to a subspace with more concentrated energy through a layered learnable orthogonal transformation module, and feature channels are clustered and grouped; different bit widths are distributed to the coefficients based on the importance degree through a non-uniform quantization module; dynamically adjusting the fragmentation size and the fragmentation sending rhythm through an RTT adaptive scheduling module according to the round-trip delay and the jitter degree; recording an accumulated reconstruction error through an error tracking compensation module and carrying out residual encoding; and periodically transmitting the anchoring layer through the key frame synchronization module to eliminate accumulated errors. According to the method, the robustness, convergence stability and communication efficiency of distributed training in a complex network environment are remarkably improved.

A data processing method, applied to the field of artificial intelligence, and comprising: acquiring first data, wherein the first data is obtained by means of a machine learning model on the basis of first input data; performing non-uniform quantization processing on the first data to obtain first compressed data, and storing the first compressed data in a memory; and reading the first compressed data from the memory, and performing inverse quantization processing corresponding to the non-uniform quantization processing on the first compressed data to obtain second data, wherein the second data and second input data are used for being input into the machine learning model, and the second input data is data input into the machine learning model after the first input data. According to the present application, non-uniform quantization is performed on data on the basis of non-uniform distribution characteristics of the data, so that the distribution characteristics of the data can be better met, and a quantization result having a smaller overall error or average error can be obtained, thereby improving the processing precision of models.

The invention discloses a polling arbitration method and a polling arbitration system for dynamically updating weight by an EWMA algorithm, and the method comprises the following steps: monitoring and receiving a request signal sent by each AHB bus channel on a matrix bus in real time; taking a preset arbitration period as a time window, collecting the current request access times of each AHB bus channel, reading the historical request times stored in the last period, calculating the request variation of each channel, and performing uniform quantization; and on the basis of the historical weight value and the quantized value, an exponentially weighted moving average value is calculated through an EWMA algorithm. When the method is used, the actual weight value of each AHB channel can be adaptively and dynamically adjusted every fixed arbitration period number, resource allocation and system fairness are improved, and it is ensured that an arbiter can more accurately reflect the actual demand of each AHB channel on a matrix bus.

The invention relates to the technical field of data processing compression and machine learning, and discloses a non-stationary data quantitative processing method based on dynamic volatility perception, which comprises the following steps of: continuously tracking real-time volatility statistical measurement of a photovoltaic data stream by adopting a Welford algorithm, and parallelly calculating a first-order differential sequence of the data stream to capture time dynamic characteristics; constructing a non-uniform quantizer with learnable parameters, setting a non-uniform input threshold by using a quantile initialization strategy, and mapping a non-uniform interval into a fixed and uniform integer output in combination with a non-uniform-to-uniform quantization mechanism; in order to guide learning of quantization parameters, a dynamic composite loss function based on volatility adaptive Beta divergence and time gradient consistency is constructed, the function comprises a Beta divergence distribution loss item dynamically controlled by real-time volatility, and the method can adapt to non-stationary characteristics of energy data such as photovoltaic power and the like. The robustness is remarkably improved while the high signal-to-noise ratio is guaranteed, and the method is suitable for edge computing scenes with limited resources.

The data processing method is applied to the field of artificial intelligence and comprises the steps that first data are obtained, and the first data are obtained through a machine learning model according to first input data; performing non-uniform quantization processing on the first data to obtain first compressed data, and storing the first compressed data to a memory; and reading the first compressed data from the memory, and performing inverse quantization processing corresponding to the non-uniform quantization processing on the first compressed data to obtain second data, the second data and second input data being used for being input into the machine learning model, the second input data is data input into the machine learning model after the first input data. According to the method, non-uniform quantization is carried out on the data based on the non-uniform distribution characteristics of the data, the distribution characteristics of the data can be better met, the quantization result with smaller overall error or average error is obtained, and thus the processing precision of the model is improved.

This invention discloses a lightweight bearing fault diagnosis method based on uniform quantization and adversarial distillation. Employing knowledge distillation and adversarial learning, the student model learns valuable information by studying soft labels provided by the teacher network. With the help of a well-learned teacher network, even a small student model can achieve diagnostic accuracy similar to deep networks. Simultaneously, combining uniform quantization with the distillation process significantly reduces the size of the student model. The method is computationally simple and effective, making lightweight networks more suitable for practical devices with limited computing and storage resources. This enables real-time bearing fault diagnosis on smaller devices such as mobile phones and embedded devices.

The invention belongs to the technical field of voice signal quantization, and particularly relates to a power enterprise smart phone customer service interaction method based on artificial intelligence, which comprises the following steps: acquiring a voice frame in real time and converting the voice frame into a frequency domain, extracting a main masking tone and a total masking threshold value, and combining with Bark sub-band weighted fusion to obtain a main masking tone and a total masking threshold value; and generating the comprehensive perception masking intensity reflecting the voice masking capability. And dynamically adjusting the sensitivity weight and the quantization relaxation coefficient of each section of the A-law thirteen-line quantizer based on the intensity, adaptively adjusting the range of each section and the quantization step length, and realizing non-uniform quantization of perception driving on the premise of keeping the input range compatible with the G.711 standard. When voice energy is concentrated, quantization is relaxed to improve compression efficiency, and when voice is sensitive, fine quantization is maintained to protect sharpness. According to the invention, on the basis of not changing the decoding compatibility, both the subjective voice quality and the coding efficiency are considered.

The invention discloses a visual converter post-training quantification method based on group awareness cooperation, and relates to the technical field of visual converters, and the method comprises the following steps: inputting a pre-training visual converter model, preparing a calibration data set, and initializing quantification hyper-parameters; calculating an original quantization parameter of each channel, constructing a feature vector after standardization, and calculating an intra-group unified quantization parameter and implementing uniform quantization by clustering dynamic grouping channels; a trainable parameter is introduced to the Softmax activation layer, a continuous displacement factor is generated through a constraint function, and nonlinear quantization is executed in combination with KL divergence optimization; a quantization error is calculated, and mean value compensation is carried out according to dimension characteristics of a convolutional layer or a full connection layer; a continuous displacement factor is pre-calculated as a fixed-point integer, and integer shift is used to replace exponential operation in inverse quantization. Through a dynamic perception grouping quantization mechanism, the problem that activation values are distributed unevenly after middle-layer normalization of the visual converter is solved.

The invention relates to an unmanned aerial vehicle physical layer secret key secret communication method based on an orthogonal time-frequency space. Comprising the following steps: establishing an unmanned aerial vehicle air-ground communication system consisting of a multi-antenna unmanned aerial vehicle base station, ground single-antenna legal users and ground passive eavesdropping illegal users; obtaining channel estimation vectors of all dominant paths by adopting a three-dimensional structured orthogonal matching pursuit algorithm; the method comprises the steps of obtaining a channel characteristic value of a communication channel through a path gain self-adaptive weighted moving average coding method, carrying out double-bit uniform quantization, carrying out Winnow key negotiation based on dynamic grouping and carrying out SHA-2 hash function confidentiality enhancement, and thus obtaining a physical layer key. According to the method, the channel dimension is converted through the orthogonal time-frequency space technology, the influence of Doppler frequency shift is effectively resisted, the channel reciprocity is improved, the key inconsistency rate is reduced, and the safety and stability of a communication system can be remarkably improved.

The invention relates to the technical field of data acquisition and calibration, and discloses a multi-channel data synchronous acquisition system and method.The method comprises the steps that when a testing machine is in a no-load state, multi-channel electromagnetic interference reference parameters are obtained, a three-dimensional interference baseline matrix Q is established, a time domain phase compensation weight coefficient K is generated, and a dynamic load topological coordinate system LS-D is constructed; obtaining a space-time reference deviation rate matrix phi; in an on-load state, based on the matrix, interference suppression and phase mismatch compensation are carried out, a chaotic feature code C is generated and optimized K is generated, an original data stream is collected, spatial-temporal trajectory compensation calibration and data calibration are carried out, non-uniform quantization cutting and recombination are carried out, and an anti-interference synchronization signal is output; according to the method, the problems of interference, phase mismatch and the like in the multi-channel data synchronous acquisition process are effectively solved, and the data acquisition precision and reliability are improved.

The invention relates to a task-driven MIMO image processing method and device based on machine consciousness, computer equipment, a computer readable storage medium and a computer program product, and the method comprises the steps: obtaining image data which comprises a sample image and annotation data; performing feature encoding processing on the sample image through an encoder to obtain a target feature, and performing dimension reforming on the target feature to obtain a target number of paths of MIMO data streams; performing uniform quantization based on a preset bit number to obtain a communication baseband symbol corresponding to the bit stream; transmitting through a communication channel, and receiving a communication baseband symbol through a receiving end to obtain a target recovery feature; performing feature decoding and task prediction through a decoder and a task processing module to obtain prediction data; training is carried out based on the prediction data and the annotation data, a trained image processing model is obtained, and deep coupling of an actual digital MIMO communication system and an intelligent agent perception task is achieved.

The invention relates to a channel estimation method and system based on a storage and calculation integrated device, processing equipment and a storage medium, and the method comprises the steps: obtaining a pilot frequency structure matrix and a probability graph model of a to-be-measured channel estimation task, and obtaining a channel estimation result according to a pilot frequency signal observed by a receiving end and a variance sum of channel uniform quantization; determining a state value initial probability of each variable node in the probabilistic graph model; constructing a node state matrix corresponding to each undirected edge in the probability graph model; constructing a driving vector corresponding to each undirected edge in the probability graph model; based on the node state matrix corresponding to each undirected edge, storage and calculation all-in-one device configuration is carried out, and a storage and calculation all-in-one device sub-array corresponding to each undirected edge is obtained; and determining a continuous channel estimation result in the to-be-measured channel estimation task according to the storage and calculation all-in-one device sub-array, the driving vector and the confidence coefficient of message transmission between each variable node and each check node connected through the undirected edge, and the method can be widely applied to the technical field of signal processing.

The application discloses a RIS codebook determination method, device and equipment, the method comprises: obtaining the initial fingerprint database corresponding to the initial RIS codebook set;Based on the RSS information entropy of the non-uniform quantization of the RSS data of each reference node, determine the candidate RIS codebook set in the initial RIS codebook set;Based on genetic simulated annealing algorithm, determine the target RIS codebook set from the candidate RIS codebook set, each RIS codebook in the target RIS codebook set is used to determine the position information of the user equipment in the positioning area. The application determines the final target RIS codebook set by introducing a two-stage RIS codebook selection, and effectively avoids introducing RIS configuration with small RSS difference, so that when positioning is carried out by using the target RIS codebook set, multiple information interaction and RSS measurement of AP and UE are not required, and the positioning accuracy and positioning time can be effectively improved.

The embodiments of the present application disclose an image data processing method, apparatus, computer device, and storage medium. The method includes: determining a first variable value of a pixel point on a first color dimension of a target color space based on the pixel value of the pixel point in the target color space within the image data; performing uniform quantization processing on the first variable value based on a quantization window to obtain m quantization sets; determining a sliding step of a sliding window based on the window size of the quantization window, and traversing the m quantization sets based on the sliding window and the sliding step to obtain n aggregation sets; selecting an aggregation set that meets the dominant color extraction condition from the n aggregation sets, determining a target pixel point in the selected aggregation set, obtaining a second variable value on a second color dimension of the target color space, and determining the dominant color of the image data based on the second variable value and the first variable value of the target pixel point. By using the embodiments of the present application, the accuracy of dominant color extraction can be improved.

The present invention discloses an adaptive non-uniform quantization decoding method, belonging to the field of electronic communication technology, which includes: obtaining the variable node initialization message and the received message, performing the quantization decoding of the first iteration on the received message by using the variable node initialization message, and obtaining the first iteration decision codeword matrix composed of the first iteration decision codewords; performing discrimination by using the above matrix, if the preset iteration end requirement is satisfied, outputting the first iteration decision codeword, if not, calculating the adaptive quantization parameter according to the first iteration decision codeword and performing non-uniform quantization on the message transmitted by the variable node by using it, then increasing the iteration number by one and performing the quantization decoding of the next iteration on the previous iteration decision codeword by using the message transmitted by the variable node to the check node in the previous iteration until the preset iteration end requirement is satisfied, outputting the decision codeword, and completing the decoding; improving the problem that the traditional method has a low adaptability to the rapid change of parameters in the decoding.

The application belongs to the technical field of optical fiber communication, and particularly relates to a low-complexity real-time nonlinear compensation method for a short-distance optical fiber link system. The application adopts Volterra algorithm to realize nonlinear damage compensation. The algorithm is regarded as an adaptive updating tap filter, and is divided into a linear term finite impulse response filter calculation module, a nonlinear term finite impulse response filter calculation module, an error calculation module and a tap coefficient updating module. Further, the low-complexity processing of the nonlinear compensation is realized through pruning and non-uniform quantization. The application solves the nonlinear distortion problem caused by devices and transmission in the intensity modulation direct detection system, and is simpler to realize and has greatly reduced complexity, thereby saving power consumption more. The application has been verified on an FPGA development board, and provides a basis for the hardware implementation of the nonlinear compensation and lays a foundation for the further development of future optical fiber communication systems.

The present disclosure provides an end-to-end joint optimization method for dynamic neural radiance field representation and compression, which comprises: representing a dynamic neural radiance field by using a coefficient feature grid and a base feature grid, compensating an error area of the dynamic neural radiance field by using a residual feature grid and the coefficient feature grid, and determining a represented dynamic neural radiance field; performing model training processing and analog quantization operation processing on the represented dynamic neural radiance field; performing optimization processing on the dynamic neural radiance field subjected to the analog quantization operation processing according to a predicted compression data amount of the represented dynamic neural radiance field and a preset loss function, and determining a trained dynamic neural radiance field; performing uniform quantization processing and encoding processing on the trained dynamic neural radiance field, and determining an actual compression data amount of the dynamic neural radiance field. Through the present disclosure, the dynamic neural radiance field is efficiently modeled and compressed, the compression rate of the dynamic neural radiance field is improved, and the distortion degree after restoration is reduced.

This invention relates to a method for secure physical layer communication for unmanned aerial vehicles (UAVs) based on orthogonal time-frequency space. The method includes the following steps: establishing an UAV air-to-ground communication system consisting of a multi-antenna UAV base station, a legitimate ground-based user with a single antenna, and an unauthorized ground-based user passively eavesdropping; obtaining channel estimation vectors for all dominant paths using a three-dimensional structured orthogonal matching pursuit algorithm; obtaining channel eigenvalues ​​of the communication channel using a path gain adaptive weighted moving average coding method; performing double-bit uniform quantization; and enhancing security through Winnow key negotiation based on dynamic grouping and SHA-2 hash function, thereby obtaining the physical layer key. This invention effectively combats the effects of Doppler shift by transforming the channel dimension using orthogonal time-frequency space technology, improving channel reciprocity, reducing key inconsistency rate, and significantly improving the security and stability of the communication system.

The application provides a precision-sensitive fine-grained recognition method based on Posit-driven symbolic pattern shaping super-dimensional calculation, and relates to the technical field of digital integrated circuits and artificial intelligence accelerator design. The pre-activation GAP semantic features of the frozen pre-training CNN are taken as input, and prototype Hamming matching is adopted to complete decision. Under the constraint of target deployment bit width, Posit non-uniform quantization is introduced to match the semantic feature distribution of the center, reduce the low bit width quantization error, and improve the low bit width precision; meanwhile, through the learning bipolar projection of quantization perception, the sign margin and dimension contribution are directly shaped, the Hamming separation between classes is expanded, and the reachable precision saturation level is raised. Since the super-dimensional calculation has the class number decoupling characteristic, when the number of classes is expanded, only the on-chip storage resources and the number of parallel comparisons need to be linearly increased, which has better energy efficiency and resource scaling characteristics compared with the traditional fully connected classification head, and significantly improves the hardware implementation friendliness on FPGA or special accelerators.

The invention discloses a fair resource management method and system for multi-user shared large language model reasoning, and the method is characterized in that the method employs a cold data recognition elimination mechanism and a fair cache distribution mechanism, carries out the unified quantification of the resource consumption of a request in GPU calculation, decoding and cache transmission, and carries out the fair scheduling based on the accumulated CPI value of each user; the system comprises a user request management module, a CPI fair scheduling module, a model reasoning execution module, a resource monitoring module and a cache management module. Compared with the prior art, the method has the advantages that users can share computing and caching resources fairly, efficient operation performance of the system is guaranteed, users with high cache hit rate are prevented from excessively occupying GPU execution opportunities due to low apparent computing overhead, the overall reasoning efficiency and the cache hit rate of the system are remarkably improved, and the user experience is improved. Unified and fair distribution of computing resources and cache resources is achieved, the effect is prominent especially in a multi-tenant high-load scene, and the method has good application prospects and commercial development value.

The embodiment of the invention provides an object detection scheme based on a lookup table, and relates to the technical field of computer vision, and the method comprises the steps: carrying out the unified quantification of weight parameters of convolution kernels of all convolution layers in a pre-trained object detection model and an activation value of a generated feature map; creating a two-dimensional multiplication lookup table; inputting a to-be-detected image into the object detection model, and when a calculation process enters a target convolutional layer, obtaining a feature map corresponding to the to-be-detected image after unified quantization and a convolution kernel in the target convolutional layer after quantization; aiming at each element in the feature map, regarding the element as a first multiplier, and regarding the corresponding element of the element in the convolution kernel as a second multiplier; inputting the first multiplier and the second multiplier into a two-dimensional multiplication lookup table to obtain binary intermediate products, performing shift zero filling on the binary intermediate products, and then performing summation to obtain a target product; and generating a detection tensor based on each obtained target product. Through the object detection method, the detection efficiency of the object in the image can be improved.

This invention relates to the field of quantum computing technology, and particularly to a method, system, computer device, and storage medium for optimizing the fidelity of quantum circuits. The method includes: acquiring the hardware parameters of the quantum circuit to be executed and the quantum processing unit cluster; standardizing the quantum circuit, analyzing the type of quantum gate and its active bits, and standardizing the multi-bit gates; constructing a unified fidelity cost model based on the quantum circuit and hardware parameters, and quantizing the cost into an additive cost; spatially partitioning the quantum circuit based on the coupling relationship between the quantum bits, and temporally partitioning the quantum circuit based on the quantum gate sequence to form a set of spatiotemporal sub-blocks; pre-calculating the partitioning cost for each spatiotemporal sub-block, and determining the candidate execution mode for each spatiotemporal sub-block accordingly; using a multi-objective global optimization algorithm to globally optimize the allocation scheme to obtain a set of optimal cooperative execution schemes; and outputting the quantum circuit mapping results.

The invention discloses a non-uniform quantization method and device based on density sensing clustering, equipment and a storage medium, and relates to the technical field of network communication, and the method comprises the steps: obtaining a tap coefficient of an equalizer; performing clustering division on the tap coefficient according to a preset quantization bit width to obtain a plurality of target clusters; determining a quantization parameter corresponding to the target cluster according to the dynamic range of the intra-cluster coefficient in the target cluster, wherein the quantization parameter comprises a scaling factor and a zero offset parameter; and independently and uniformly quantizing the tap coefficient in the target cluster according to the quantization parameter to obtain a recovery coefficient. According to the invention, the coefficient is adaptively divided into a plurality of clusters based on density sensing clustering, and uniform quantization is independently performed on each cluster, so that high-precision quantization of the equalizer coefficient under extremely low bits is completed, and the storage efficiency of coefficient quantization and the performance of the equalizer are improved.

The application discloses a mixed multi-resolution channel state information feedback method and device, relates to the technical field of channel state information feedback in a large-scale multiple-input multiple-output system, and comprises the following steps: obtaining an angle-time delay domain channel state information matrix, performing feature embedding on the angle-time delay domain channel state information matrix through an initial feedforward network to obtain initial features; inputting the initial features into a multi-resolution convolution model and a multi-head sharpening attention model in parallel, respectively extracting multi-scale local features and long-distance dependent features, adaptively weighting and fusing the multi-scale local features and the long-distance dependent features through learnable weights to obtain fused features; converting the fused features into low-dimensional feedback code words through a feedforward network, an activation function and a compression mapping, and then adopting local uniform quantization segmentation processing to generate quantized code words, and feeding back the quantized code words to a receiving end to reconstruct channel state information. The application reduces the calculation complexity of channel feedback and quantization loss, and can realize low-latency and high-fidelity channel reconstruction on a resource-limited device.

This invention discloses a simple intermittent quantization control method for Chua's circuit systems. This method first designs an indirect signal-based estimation algorithm to estimate unknown states and disturbances in the network system. Then, two threshold functions are designed to construct an aperiodic intermittent control strategy. A uniform quantization controller quantizes the system's control inputs to reduce the transmission of redundant control signals. When applied to Chua's circuit systems, this method achieves excellent unknown state and disturbance estimation results and effectively conserves system control resources.

The invention belongs to the technical field of chemical engineering, artificial intelligence and model compression, and discloses a uniform quantification method for a large model in the chemical engineering field. On the premise that the expression capability of a model function is not changed, orthogonal equivalent transformation is introduced, combined smoothing processing is carried out on the weight and activation in a linear layer, and activation distribution approximately obeys uniform distribution in a quantization interval by constructing an orthogonal transformation matrix and systematically remodeling the activation distribution, so that the low-bit uniform quantization error is remarkably reduced. And carrying out joint quantization on the smoothed weight and activation by adopting a uniform low-bit uniform quantizer, further introducing a small number of learnable global orthogonal parameters, and carrying out lightweight optimization on module output errors before and after quantization. According to the method, quantitative perception training or large-scale fine tuning is not needed, calculation and storage overhead is low, engineering implementation is simple, inference performance close to full precision can be kept under the 4-bit or 6-bit quantization condition, and video memory occupation and energy consumption in the model deployment process are remarkably reduced.

The invention provides a post-training quantification method for a medical image segmentation basic model, and belongs to the field of medical images. The method comprises the following steps: acquiring a to-be-quantized medical image segmentation basic model and a pre-training weight; constructing a medical segmentation image standard data set, and performing preprocessing; determining a target bit width, and selecting uniform quantization as a basic quantization mode; obtaining a new weight matrix; alternately solving an integer coding matrix and a scaling factor; and constructing a quantized medical image segmentation basic model, and performing reasoning on the medical segmentation image standard data set. The technical problem that the performance of an existing post-training quantification method is seriously degraded under a low bit width due to the fact that the weight distribution of a medical image segmentation basic model is highly heterogeneous in various medical imaging modes and downstream segmentation tasks is solved; and additional calculation or memory overhead is not introduced in the reasoning stage, so that the deployment feasibility of the quantized medical image segmentation basic model in resource-limited clinical / edge equipment is improved.