360 results about "Quantitative model" patented technology

The invention discloses an intelligent management system for energy consumption optimization and fault self-diagnosis of cleaning equipment, and relates to the field of intelligent maintenance of the cleaning equipment, and the system comprises the steps: obtaining three groups of core parameters, i.e., a historical vibration spectrum, a motor current harmonic component and a bearing temperature gradient, constructing a dynamic failure mode knowledge graph, performing time sequence correlation analysis on the historical fault data to obtain failure mode analysis data; establishing a multi-dimensional analysis platform, identifying a high-risk component, and updating a fault threshold value; introducing a service time attenuation factor and a working condition correction coefficient, establishing an aging degree quantitative model, and calculating an aging coefficient; and constructing and developing an energy consumption-reliability joint optimization module, and adjusting equipment operation parameters. The method has the advantages that the dynamic knowledge graph and the time sequence analysis model are constructed by integrating multi-source sensor data, precise diagnosis and self-adaptive threshold adjustment of the coupling fault of the cleaning equipment are achieved, aging evaluation and task scheduling optimization are combined, the energy consumption efficiency is improved, and the maintenance cost is reduced.

The invention provides a titanium alloy bar visual detection method based on image segmentation. The method comprises the following steps: carrying out pretreatment and feature enhancement operation on an acquired titanium alloy bar surface image; constructing an adaptive segmentation algorithm based on region growth to carry out defect region segmentation on the surface image of the titanium alloy bar; performing defect feature extraction on the segmented defect region; a defect classification model is constructed, and titanium alloy bar defect types are classified based on defect features; performing quantitative analysis on the defect severity through a defect severity quantitative model; establishing a defect causal chain model, and carrying out defect traceability analysis and causal derivation; after defect traceability analysis is completed, fusion analysis is carried out on the multi-scale features, and decision suggestions are given. According to the method, the surface defects of the titanium alloy bar can be efficiently and accurately recognized and classified, the accuracy and robustness of defect detection are improved, and the defects of a traditional detection method in the aspects of efficiency, recognition capability, quantitative analysis and the like are effectively overcome.

The invention relates to the technical field of industrial Internet of Things, in particular to a machine equipment online state monitoring and fault diagnosis system, which comprises the following steps of: acquiring multi-source heterogeneous sensing data through an edge computing node deployed on an equipment body, performing adaptive noise filtering and feature dimension reduction processing on original data, and acquiring multi-source heterogeneous sensing data; outputting a standardized equipment state vector set; inputting the equipment state vector set into a dynamic knowledge graph engine, constructing a fault evolution network comprising space-time correlation characteristics based on an equipment operation entropy change quantification model, and generating a graph node connection relationship with a weight coefficient; and inputting the fault evolution network into a migration reinforcement learning module, and outputting a diagnosis decision set comprising a fault type, a severity degree and an evolution path through knowledge migration of a cross-device fault mode. According to the method, the problems of edge redundancy and single feature expression in traditional rule-based atlas construction are effectively avoided, and the structuring ability and physical traceability of fault recognition are improved.

The invention discloses a road safety early warning method and system based on a mixed precision quantification visual large model, and the method comprises the steps: collecting road traffic safety videos and pictures, and carrying out the preprocessing, data enhancement and marking, thereby forming a diversified data set; a pre-trained visual large model is selected as a teacher model, after fine tuning, output layer and middle layer knowledge is extracted, key features are weighted, and meanwhile, a lightweight neural network is taken as a student model, same input is received, and prediction and middle feature maps are output. And inputting data into the two models and the student model to carry out mixing precision quantification forward propagation, constructing a total loss function containing tasks, knowledge distillation and quantification learning loss, and updating parameters through back propagation. And after training is completed, exporting a quantitative model, and deploying the quantitative model to an edge computing platform to realize safety early warning. The lightweight model can realize rapid reasoning on edge equipment such as a vehicle-mounted road side, and the problem that performance and efficiency are difficult to consider in a traditional model compression method is solved.

The invention relates to a geological engineering double-dessert comprehensive evaluation method, device and equipment and a medium, and belongs to the technical field of oil and gas field development, and the geological engineering double-dessert comprehensive evaluation method comprises the following steps: determining evaluation parameters of geological desserts and engineering desserts based on obtained data of a tight sandstone gas reservoir; taking the open flow capacity of the fractured well in the tight sandstone gas reservoir as a reference sequence and the evaluation parameters as a comparison sequence, determining a geological sweet spot index and an engineering sweet spot index by adopting a grey correlation method, and determining a comprehensive sweet spot index of the geological engineering double sweet spots based on the geological sweet spot index and the engineering sweet spot index; the quantitative model is constructed based on the open flow capacity and the comprehensive sweet spot index, the yield of the to-be-predicted fractured well is predicted based on the quantitative model, the perforation position of the to-be-predicted fractured well is optimized, and the mining efficiency of the tight sandstone reservoir is improved.

The invention discloses a forest-town junction region fire dynamic early warning method based on multi-source data assimilation, and the method comprises the following steps: inputting collected multi-source observation data into a fire risk prediction model for training, and outputting a fire risk grid map covering the whole forest-town junction region for monitoring a high-risk building group; constructing a fire spreading numerical model covering multi-material, multi-scale and multi-physical mechanism coupling; a fire propagation characteristic quantitative model is constructed, and the building heat influence is evaluated; and constructing a data assimilation and fire dynamic prediction model, and dynamically predicting and outputting fire spreading parameters and a spreading trend chart. According to the method, a forest-building-oriented complex scene is constructed, and propagation behaviors of flames under multiple mechanisms of convection, radiation and conduction are effectively described; through the modes of risk identification, dynamic simulation, data assimilation, visual early warning and the like, the precision, response speed and engineering applicability of forest-town boundary region fire prediction are remarkably improved.

The invention provides a target recognition model reasoning optimization method and device, and the method comprises the steps: firstly carrying out the structural analysis and sensitivity evaluation of a pre-training model, extracting the structural features of each network layer, activating the distribution features, carrying out the quantitative sensitivity scoring, and constructing a data set reflecting the hierarchical features and fault-tolerant capability; and querying a quantitative configuration knowledge base based on the data set to generate a heterogeneous quantitative strategy. Layered low-bit quantization is executed according to the strategy, and a layered weighted loss function is introduced to carry out quantization perception training, so that precision loss caused by bit width compression is effectively compensated. According to the method, through hierarchical heterogeneous quantification, the model recognition precision is preserved to the maximum extent while high compression ratio and reasoning acceleration are achieved, and particularly, the performance of a high-sensitivity layer is protected. The generated heterogeneous quantitative model remarkably reduces memory occupation and power consumption, is suitable for an edge hardware platform with limited resources, forms a set of complete automatic process from analysis and configuration to training compensation, and has good universality and engineering practical value.

The invention provides a temperature and humidity compensation method and system for a gas sensor, and relates to the technical field of temperature and humidity compensation methods, and the method comprises the steps: obtaining the temperature and humidity of a space where the gas sensor is located and the time sequence data of the gas concentration output by the gas sensor at the same time, and obtaining a temperature input coefficient and a humidity input coefficient; establishing a temperature and humidity cross influence quantitative model to obtain a gas concentration error of the gas sensor caused by temperature and humidity changes; and filtering the gas concentration data output by the sensor before compensation, constructing a weighted comprehensive index of temperature and humidity changes, and compensating the filtered gas concentration data by adopting a dynamic switching compensation mode. And establishing a temperature and humidity cross influence quantitative model, inputting the temperature input coefficient and the humidity input coefficient into the established temperature and humidity cross influence quantitative model to obtain a gas concentration error, and adopting a dynamic switching compensation mode based on the gas concentration error according to the change of the environment temperature and humidity.

The invention belongs to the technical field of computer artificial intelligence, and relates to a citrus intelligent planting management-oriented large model field quantification and adaptive model deployment method, which comprises the following steps of: firstly, constructing a large model calibration data set and carrying out preprocessing, and inputting a Transform model to execute forward reasoning; then carrying out left multiplication rotation on a weight matrix of a to-be-quantized layer of the model and right multiplication rotation on an activation matrix, executing GPTQ quantization, calculating a dynamic adaptive smoothing factor of each channel of the activation matrix, executing normalization and symmetric quantization according to hidden dimension grouping, and generating a quantized Transform model; then obtaining citrus industry text data to construct a fine tuning instruction set, and performing supervised training on the quantitative model to generate a full-quantitative and full-precision model; and finally, by training a task complexity classifier, selecting a full-quantization or full-precision model as a target deployment model according to a task complexity level, so that efficient quantization and intelligent task adaptation of the model are realized.

The invention discloses a source-load interaction method based on an electricity utilization information acquisition system, and relates to the field of power distribution network operation and optimization. The method comprises the following four steps: S1, power consumption data acquisition: dynamically acquiring user power consumption, distributed energy output and power grid state data through intelligent equipment, and ensuring real-time accuracy; s2, load characteristic analysis: constructing a quantitative model to analyze characteristics such as a user load elastic coefficient and adjustable potential, establishing a supply and demand balance evaluation system in combination with distributed energy fluctuation, and determining peak and valley time periods and gaps; s3, generating an interactive strategy, and generating and issuing dynamic strategies such as electricity price excitation, load regulation and energy scheduling based on an analysis result; and S4, executing and feeding back the strategy, monitoring the execution effect in real time, and optimizing the strategy through a closed loop mechanism. According to the method, the problems of staticization, single mode and low efficiency of a traditional method are solved, the power grid supply and demand balance capability, the renewable energy consumption rate and the strategy accuracy are improved, and low-carbon and intelligent operation of the power distribution network is supported.

The invention provides a method for predicting the quality of a sparse small sample industrial process based on a stacked network model. The method comprises the following steps: collecting end point quality report data of an industrial production process; performing hierarchical processing on the acquired data according to the missing rate, and removing abnormal data in combination with a quartile method and production experience; generating a synthetic data expansion small sample data set by adopting a conditional generative adversarial network; obtaining a first-layer basic model based on an accumulated contribution rate screening method of an SHAP value; constructing a first layer of a stacked integrated learning model and adjusting hyper-parameters by using Bayesian optimization; constructing a Ridge meta learning device to integrate the output of the basic model and constructing a second-layer network; a six-fold cross validation training model is adopted; predicting performance through a multi-index quantitative model based on the test set; and verifying the prediction precision of the end-point phosphorus content and the temperature by using real converter production data. The method can realize high-precision prediction of the end point quality index of the complex industrial generation process, and is beneficial to ensuring the product quality and improving the production efficiency.

The invention discloses a power transmission line tower footing bare soil vegetation restoration capability comprehensive evaluation method, and relates to the technical field of vegetation restoration capability evaluation, and the method comprises the steps: building an evaluation index system, building a quantitative model, carrying out data collection, and carrying out calculation and grading. According to the method, a comprehensive evaluation index system is constructed, electromagnetic radiation related indexes and microtopography related indexes are covered, the actual vegetation restoration situation can be reflected more accurately, the influence coefficient is measured and calculated in detail, the influence of electromagnetic radiation and microtopography changes on vegetation is fully considered, a basis is provided for accurate evaluation, and the vegetation restoration efficiency is improved. The problems of inaccurate and incomplete evaluation are effectively solved, quantitative model construction and optimization are performed, time factors are introduced to dynamically evaluate the vegetation restoration capability, problems can be found in time, a restoration strategy can be adjusted, accurate grade division is performed, reasonable selection of vegetation varieties is facilitated, proper restoration measures are adopted, the vegetation restoration success rate is increased, and the vegetation restoration efficiency is improved. The ecological stability of the tower footing area is enhanced.

The invention is suitable for the field of industrial safety monitoring, and provides an industrial control safety monitoring method and system, and the method comprises the steps: collecting the network flow data of an industrial control network in real time based on an industrial protocol analysis engine; synchronously acquiring operation state parameters of the key equipment by adopting an edge computing node; constructing a coupling relation quantitative model of the network flow and the equipment operation state, and calculating a dynamic correlation coefficient between the network flow and the equipment operation state through the coupling relation quantitative model; constructing reference correlation coefficient distribution in a normal production state based on historical data, and dynamically updating a safety threshold value; when the dynamic correlation coefficient exceeds a safety threshold value, triggering a multi-stage early warning mechanism; according to the method and the system, deep fusion monitoring of the industrial control network and the equipment operation state is realized, and the limitation of traditional split type monitoring is broken through. Through dynamic quantification of a coupling relationship, adaptive threshold adjustment and multi-level early warning, the timeliness and accuracy of abnormal risk identification are significantly improved, and network attacks and equipment fault hazards are effectively reduced.

The invention discloses a damping thermal insulation material performance evaluation method and system for a battery pack, and relates to the technical field of battery pack thermal management, and the method comprises the steps: extracting a calibration sample set, applying transient thermal excitation, and collecting an infrared thermal image sequence to extract thermal response features; performing industrial CT scanning to obtain defect quantitative indexes, and forming a calibration data set with the thermal response characteristics; establishing a quantitative correlation model based on the calibration data set, and determining a batch quality threshold according to model output distribution; executing the same excitation and collection on an online sample to be detected to extract the characteristics of the sample, inputting the model to obtain a defect quantitative prediction value, comparing the defect quantitative prediction value with a batch quality threshold, and outputting an evaluation result; when a preset period or a trigger condition is met, a new sample is supplemented and extracted, and newly added calibration data is obtained to update the model and the threshold value. According to the method, accurate material performance evaluation is provided by establishing a quantitative model of thermal response characteristics and defect quantitative indexes, and evaluation consistency and stability are enhanced through a dynamic updating mechanism.

The invention relates to the technical field of traffic transportation environment engineering and carbon emission evaluation, in particular to a construction method and equipment of a plateau gradient zone vehicle CO2 emission factor calculation model and a medium. Through the method, a route-level CO2 emission factor distribution map is obtained, and different quantitative calculation models are formed for different altitude intervals. And a specific and quantitative design basis is provided for low-carbon highway design. It is revealed for the first time that the CO2 emission rate of the plateau gradient zone has a remarkable altitude segmentation effect through empirical data, and a method for constructing a calculation model in a partition mode is innovatively provided. According to the method, the defects that a traditional single model is poor in adaptability and misaccurate in prediction in a complex plateau environment are overcome, and differentiated quantitative models (such as a low-altitude quadratic function and a high-altitude S-type function) are established for different altitude intervals; the finally formed partition model provides a core theoretical tool and decision basis for plateau highway carbon emission accurate accounting, low-carbon route optimization and green traffic construction.

The invention relates to the technical field of spectral analysis and concentration quantitative detection, and discloses an LIBS quantitative analysis method, device and system based on a thin plate spline regression algorithm. The method comprises the steps that multiple sets of sample data are obtained, and each set of sample data comprises corresponding target sample concentration, laser energy and spectral intensity; constructing a concentration quantitative model about the concentration, the laser energy and the spectral intensity based on a spline function; determining an optimal penalty parameter by adopting a cross validation method; based on the optimal penalty parameter, solving parameters of the concentration quantitative model through a minimization objective function to obtain a constructed concentration quantitative model; and performing quantitative analysis on a sample with unknown concentration by using the constructed concentration quantitative model. Thus, the output energy of the laser is incorporated into the core model, and an I-E-C nonlinear coupling relationship is constructed, so that the problem of calibration deviation caused by laser energy fluctuation is reduced, and the precision improvement of concentration quantification in a complex scene is realized.

The invention discloses a customer service full-process digitalization system and method for small and medium-sized financial institutions, and particularly relates to the technical field of customer service full-process digitalization. Through multi-modal accurate understanding and task process initialization of a customer request, a deep semantic analysis model is introduced to extract a customer intention map and a secondary intention; two key calculation indexes of a historical process tracing abnormal coefficient and an abnormal closed-loop non-achievable coefficient are introduced, double quantitative evaluation of the execution stability and the closed-loop judgment accuracy of a customer service system is achieved, and the two coefficients are organically integrated into a pseudo closed-loop deception response error index. A unified quantitative model is established to judge whether the customer service system has a high-risk false closed-loop phenomenon at present, real-time intelligent marking of a false closed-loop process is realized based on comparison of a false closed-loop deception response error index and a set threshold value, a high-risk process is accurately identified, and an early warning or re-judgment mechanism is dynamically triggered.

The invention relates to the field of gas concentration detection, and particularly discloses a real-time feedback and adaptive learning method for natural gas infrared spectrum measurement, which comprises the following steps: S1, acquiring infrared spectrum information of a natural gas body by using an infrared spectrometer, and constructing a historical sample set; s2, preprocessing the spectral data of the historical sample set; s3, selecting an optimal algorithm and a hyper-parameter by adopting XGBoost and Bayesian optimization; s4, constructing a qualitative model to identify gas types and match data; s5, calculating the similarity between a field sample and a historical sample through a Siamese network, and setting a threshold value to screen local data; s6, improving the KNN to construct a local dynamic quantitative model to predict the concentration; s7, processing low-similarity abnormal data by the global dynamic model, and improving the reliability by combining moving average and abnormal calibration; and S8, introducing reinforcement learning and online gradient descent to adjust parameters in real time to optimize the precision. According to the technical scheme, high-accuracy natural gas detection can be carried out in a complex environment.

The invention discloses an intelligent business decision support system based on multi-modal data and a knowledge graph, and relates to the technical field of data analysis. In order to solve the problems of weak data integration capability and poor model adaptability in existing business analysis, the adopted scheme comprises a multi-modal data processing module used for collecting multi-modal data and constructing a knowledge graph after preprocessing and feature extraction; the model training and optimizing module is used for taking a large-scale pre-training language model as a core, acquiring external knowledge from the atlas in combination with a retrieval enhancement generation technology, performing parameter fine tuning, optimizing a training process and generating an optimized model at the same time; the intelligent decision generation module is used for generating a preliminary suggestion based on the optimization model and generating a report through logic deviation evaluation and correction; the dynamic feedback module is used for constructing evaluation system quantitative model performance by taking the report as input and combining real-time business data; and meanwhile, user feedback is collected to form an incremental sample dynamic optimization model. The accuracy of business decision making can be improved.

The invention provides a transient voltage stability control method and system based on transmission and distribution cooperation. The method comprises the following steps: constructing a power distribution network reactive power support capability quantification model; quantizing a reactive power support capability boundary in real time by solving a power distribution network reactive power support capability quantification model, and determining a reactive power support power feasible region capable of being provided; and based on a quantitative evaluation result, dynamically adjusting a reactive power interaction value between the power transmission and distribution networks by adopting an improved droop control strategy, and realizing power transmission network transient voltage stability control based on active support of the power distribution network. Power distribution is carried out according to the dynamic reactive power margin of the reactive power regulation equipment by considering the transient voltage support requirement of the power transmission network and the reactive power transmission loss of the power distribution network; and obtaining a reactive power regulation instruction, and coordinating reactive power resources at the power distribution network side according to transient voltage stability control and power distribution of the power transmission network to realize cooperative control. Based on the method, the invention further provides a transient voltage stability control system based on transmission and distribution cooperation. The optimal transient voltage support is provided for the power transmission network.

The invention provides a model quantization implementation method, a model and computer equipment, and the method comprises the steps: obtaining a target floating point model and a training data set, and determining the quantization configuration corresponding to each neural network layer contained in the target floating point model; the quantization configuration corresponding to different neural network layers is related to the change degree of the performance of the corresponding neural network layer compared with the target floating point model after the corresponding neural network layer is quantized, and according to the quantization configuration corresponding to each neural network layer, performing pseudo quantization on parameters of the corresponding neural network layer in the target floating point model to obtain a first pseudo quantization model; and based on the target floating point model and the training data set, performing quantitative perception training on the first pseudo-quantitative model to obtain a target quantitative model, and deploying the target quantitative model to a terminal device to execute a calculation task.

The invention relates to the technical field of talent management and data analysis, in particular to a multi-source data driven scientific and technological talent ability portrait accurate matching method. The method comprises the following steps: collecting and preprocessing multi-source talent data, and mining text information by applying a natural language processing technology; constructing a talent knowledge structure graph by identifying the relationship between the key entities and the extracted entities; determining a plurality of capability dimensions, and establishing a quantitative model or rule for each dimension to measure the capability; calculating scores of all dimensions by comprehensively considering personal differences of talents and working situation factors, and generating a visual ability portrait; a demand standard is determined by analyzing the situation of a demand side, the matching degree between talents and the demand standard is calculated after the talents are screened from a talent pool, the recommended talents are ranked according to the matching degree, and related parameters and rules are optimized according to feedback of the demand side. The talent ability can be comprehensively and accurately described, efficient and accurate matching is achieved, the method adapts to dynamic changes, and talent configuration in the science and technology service field is effectively optimized.

The invention relates to an instruction-level man-machine collaborative driving method and device, and the method comprises the steps: designing an uncertainty quantitative model for the problem of hybrid uncertainty in a man-machine fusion process, and proposing a human instruction optimal approximation strategy based on safety constraints, and guaranteeing that when man-machine decisions are inconsistent, the human instruction optimal approximation strategy is not consistent with the human instruction optimal approximation strategy. Safety and human intention are considered; a human intention prediction model is constructed based on a learning and interaction model, so that the efficiency and effectiveness of man-machine fusion intelligent decision making are improved. Therefore, the problems that control right switching of a traditional cooperation mode lacks smooth transition, and the movement freedom degree of a driver is limited are solved.

The embodiment of the invention provides an embedded AI power grid waveform dynamic bit width adaptive preprocessing method and system and a medium, and the method comprises the steps: obtaining the real-time sampling data of voltage and current, and analyzing the sampling precision of the real-time sampling data; setting a sampling precision threshold, and judging whether the sampling precision is greater than the sampling precision threshold; if the sampling precision is greater than the sampling precision threshold value, performing Q-format fixed-point number conversion on the real-time sampling data, and performing normalization processing on the real-time sampling data after the Q-format fixed-point number conversion; if the real-time sampling data is smaller than or equal to the real-time sampling data, converting the real-time sampling data into normalization by adopting an LUT table look-up method; normalized sampling data is input into a quantitative model, data bit width is output, a bit width adjustment strategy is generated based on a set bit width threshold value, the data bit width is dynamically adjusted, the data preprocessing efficiency is improved through Q format fixed-point and LUT lookup methods, memory occupation is reduced, and self-adaptive preprocessing can be conducted according to collected signals of different bit widths.

The invention provides a PC (Personal Computer) end quantitative model reasoning method. The method comprises the following steps: S1, loading input information; s2, deserializing the model; s3, operator calculation is carried out; and S4, storing the data, and selecting a dump data result. The quantitative model reasoning on the PC end is used for simply reasoning the model, verifying the result of the model and obtaining the data of the network layer of each stage reasoned in the model, and the PC end reasoning is used for correcting the data, assisting the quantitative model, reasoning the float model, post-quantitative model and training the quantitative model and verifying the correctness of the data. According to the method, model reasoning is convenient and effective, model data verification is facilitated, and data information of a quantitative model is provided.

The embodiment of the invention discloses an electrical stimulation parameter determination method, device and equipment, a medium and a medical system. The method comprises the following steps: acquiring a brain three-dimensional model corresponding to a target user of which the brain is implanted with at least one stimulation electrode; determining at least one electrode contact combination corresponding to the target nucleus simulation model according to the first spatial position information of the target nucleus simulation model and the second spatial position information of the electrode simulation model; determining at least one group of candidate electrical stimulation parameters of the electrode contact combination according to the electrode contact combination and a target nucleus simulation model; determining the total power consumption of the stimulator of the candidate electrical stimulation parameters according to a preset total power consumption quantitative model of the stimulator and the candidate electrical stimulation parameters; and determining at least one group of target recommended electrical stimulation parameters from the plurality of groups of candidate electrical stimulation parameters according to the total power consumption of the plurality of stimulators. According to the technical scheme, the effect of accurately screening the recommended electrical stimulation parameters with low energy consumption and efficient stimulation effect according to the electrical stimulation power consumption is achieved.

The invention provides a deep learning model diversity data-free quantification method based on a generator. The method comprises the steps of obtaining basic feature representation of a category label of target generation data; the generator maps generation features of different levels through a normalized stream, and each layer of attention mechanism adds one mapped noise data to learn the relationship between different categories of labels to generate synthetic data; inputting the synthetic data into a full-precision model and a quantitative model at the same time, and extracting different levels of recognition features from the full-precision model; the multi-layer feature mixer obtains identification features of the same type of synthetic data at different levels and carries out linear fusion, and a fusion result is used for initializing basic feature representation; and aligning each level of the quantitative model and each level of the full-precision model, and calibrating the quantitative model by using the distribution difference output by the full-precision model and the quantitative model. According to the method, the diversity quantification requirement of the data-free quantification model is met, and the effective improvement of the model quantification effect is ensured.

The invention provides a distributed training method and device of a neural network model, a medium and a cluster, and relates to the field of artificial intelligence, in particular to the field of chips. According to the specific implementation scheme, when it is detected that the model scale of a to-be-trained model is larger than the processing scale of a distributed cluster, the model is quantized to obtain a quantized model; splitting the quantitative model into a plurality of model partitions and configuring the model partitions in each pipeline parallel group; when forward calculation is carried out, a training error value is generated through common cooperation among the pipeline parallel groups; and in the process of carrying out reverse calculation on the quantitative model according to the training error value, the model weight is updated through each pipeline parallel group according to the gradient scaling value independently maintained in the group. The parallel groups are used as granularity, each parallel group maintains the respective gradient scaling value and updates the weight in the respective model partition by using the gradient scaling value, so that the model precision and the calculation overhead can be effectively balanced, and the efficiency, the precision and the reliability of distributed training are improved.

The embodiment of the invention discloses a quantitative model degradation correction method and device, electronic equipment and a storage medium, relates to the technical field of data processing, and is convenient for reducing errors generated by a quantitative model. The method comprises the following steps: acquiring a key operator of an original model; implementing the key operator by adopting a model reasoning framework to obtain a model reasoning operator of a quantized version; registering the model reasoning operator into a deep learning training framework, and replacing a corresponding operator in the deep learning training framework with the model reasoning operator to obtain a quantitative model; injecting a fine-tuning low-rank weight into each linear layer of the quantitative model to obtain an injected quantitative model; performing fine tuning on the post-injection quantification model to obtain a fine tuning parameter result; and loading the fine tuning parameter result to the quantitative model. The method is used for degeneration correction of the quantitative model.

The invention discloses a meta-model decision-making-based intelligent prediction method for inclusion quality in an electroslag remelting process. The method comprises the following steps of: constructing a sample data set containing process and component characteristics and target variables; obtaining a first-layer basic model based on an SHAP value cumulative contribution rate screening method; a heterogeneous learner is adopted to construct a first layer structure of the stacked integrated learning model, and Bayesian is adopted to carry out adjustment and optimization; adopting logistic regression as a meta-learner to construct a second-layer structure of the stacked integrated learning model; training by adopting a five-fold cross validation strategy, and predicting the performance by using a multi-index quantitative model; and deploying the D-type inclusion prediction model in the electroslag remelting process to an actual process, collecting process parameters as input data in real time by using a multi-sensor group, performing D-type inclusion risk prediction of a corresponding heat, and outputting an inclusion risk prediction result. According to the method, the accuracy and reliability of D-type inclusion prediction can be remarkably improved.