33 results about "Model tuning" patented technology

The invention provides a large model tuning method and system based on multi-modal information and AI, and relates to the technical field of artificial intelligence, and the method comprises the steps: extracting multi-modal parameter features and interaction features, and dynamically adjusting a contribution coefficient to achieve feature migration, and obtaining a cross-modal fusion vector; aligning the modal features by adopting an adversarial mechanism to obtain domain invariant features; constructing a prompt chain based on the modal confidence, optimizing the prompt chain by using AI reinforcement learning, and iterating to converge to obtain an optimization result; and finally updating the large model parameters. According to the method, effective fusion of multi-modal information is realized, and the tuning effect and generalization ability of the man-machine interaction large model are improved.

Auto-tuning covariances associated with a set of noise models for a variety of sensor modalities and / or perception components such that the covariances are leveled respective to one another may include whitening the covariances and / or error models and determining scalars to apply to the covariances. Determining these scalars may comprise using the residuals that result from generating the set of noise model (e.g., such as may be determined as part of least squares estimation) along with the hat matrix of the process model to determine the scalars. The covariances may iteratively be updated until the scalar adjustments converge or until another end condition is met.

The invention relates to the technical field of power prediction of a power system, and discloses a sample adaptive iterative training method and system for existing model tuning, and the method comprises the steps: obtaining the net load and solar irradiance data of a power distribution network; determining a payload-irradiance response intensity indicator based on the response slope within the local window; determining reference values of a truncation mode and a response mode according to statistical distribution of the indexes, and calculating a non-truncation probability of the sample in a normal power generation state; calculating the marginal response intensity of the prediction model to the solar irradiance; constructing a critical factor for capturing an inverter action boundary and a sample training weight by using a non-truncation probability; constructing a photovoltaic bistable constraint item, and constraining marginal response intensity to approach corresponding modal reference values in different states; and finally, performing parameter updating on the model based on the weighted prediction error and the bistable constraint term. According to the method, the problem of model tuning failure caused by equipment control truncation is effectively solved, and the prediction performance of the model under different operation conditions is ensured.

To tune a machine-learning model for a deployment system, a machine-learning model training system generates multiple tuning steps for the machine-learning model and generates an accuracy loss sensitivity for each tuning step. Each of the tuning steps indicate a corresponding set of one or more parameters and hyperparameters that reduces the impact of the machine-learning model on the system performance of the deployment system. Based on the accuracy loss sensitivities, the machine-learning model training system selects a tuning step with the least impact on the accuracy of the machine-learning model and modifies the machine-learning model based on the selected tuning step. After also modifying the tuned machine-learning model based on a threshold accuracy, the machine-learning model training system provides the tuned machine-learning model to the deployment system.

A kind of H&E image ovarian cancer polyploid giant cell (PGCCs) detection method based on deep learning, including constructing specific data set, image annotation and data division, using OCDet model to carry out feature learning and optimization, and model training and evaluation.The method first establishes the H&E staining image data set containing PGCCs, then accurately labels the image and divides it into training, verification and test set.OCDet model takes CSPDarkNet as the core, combines ECA mechanism, focuses on the deep learning and re-encoding of pathological semantic features.Through the training data set, the model updates parameters through back propagation and gradient descent, optimizes to identify PGCCs features.The verification set is used for model tuning, and the test set is used for evaluating the performance of the model.The automatic detection technology of the application can help doctors improve the diagnosis efficiency and reduce the error, provide an important reference for clinical treatment and prognosis evaluation, and show significant clinical application value.

The invention discloses an intelligent processing method and system for multi-modal data of an oral broadcast video based on deep learning, and relates to the technical field of oral broadcast video data processing, and the method comprises the following steps: 1, collecting multi-modal oral broadcast data in real time, recognizing emotion types, and carrying out the data preprocessing; step 2, constructing an oral playing animation generation model, judging the synchronization degree of the mouth shape, the head action and the audio in the oral playing animation, and taking coordinated regulation measures; 3, judging the consistency of the mouth shape and the emotion intensity in the oral playing animation, and implementing an anomaly repair measure; 4, constructing a depth time sequence prediction model, evaluating the difference degree between the actual oral playing animation and the prediction result of the depth time sequence prediction model, and carrying out video optimization and model tuning; the problems that when mouth shape supplement and animation filling are carried out by AI, precise synchronization of mouth shapes, expressions, head actions and emotional fluctuations is difficult to achieve, so that visual abnormity is caused, and the reality sense and the professional degree of videos are affected are solved.

The invention discloses a self-supervised Chinese character generation method introducing dynamic convolution to enhance font style extraction, which is suitable for the fields of computer vision, font design and calligraphy education digitization, and comprises the following steps: constructing a dynamic convolution enhanced self-supervised network model which introduces a dynamic convolution kernel into the font style extraction field, and the convolution kernel parameters are dynamically adjusted to adapt to the edge and structure characteristics of the Chinese characters of different styles, so that the accurate capture of the style information is realized. According to the method, font style information extraction is enhanced through dynamic convolution, and the problems that in a traditional font generation method, a fixed convolution kernel is depended, the kernel scale and weight need to be manually adjusted, and the fixed kernel and a style extraction target are obviously separated, so that the model adjusting and optimizing process is tedious, and the cross-style generalization ability is poor are solved; therefore, the feature extraction adaptability and accuracy of different styles of Chinese characters are remarkably improved.

A method of tuning a machine learning model associated with a lithography patterning process is disclosed. More particularly, a method of using a reinforcement learning tuning engine to accurately adjust a deep learning-trained lithography model and enable global view of a training dataset is disclosed. In the reinforcement learning process, the reward can be defined as a function of global metrics derived from model predictions. Given the reward, the accuracy agent can determine actions corresponding to delta variables of the model.

The invention provides a model adjustment method and device, and relates to the technical field of communication, and the method comprises the steps: adjusting a first model parameter according to the environment information of a wireless channel, and obtaining a second model parameter, the first model parameter being a part of the model parameter of a first target model; according to the second model parameter, updating information is sent, and the updating information is used for indicating updating of the first target model. According to the scheme provided by the invention, the adaptation of the model can be realized without adjusting all model parameters, and the performance of processing the communication service by the model is improved.

The application provides a large model optimization method and system based on multi-modal information and AI, relates to the technical field of artificial intelligence, and comprises the following steps: by extracting multi-modal parameter features and interaction features, dynamically adjusting a contribution coefficient to realize feature migration and obtain a cross-modal fusion vector; adopting an antagonistic mechanism to align the modal features to obtain domain invariant features; constructing a prompt chain based on modal confidence, and optimizing the prompt chain by using AI reinforcement learning, and iterating to convergence to obtain an optimization result; and finally updating the large model parameters. The application realizes effective fusion of multi-modal information, and improves the optimization effect and generalization ability of the human-computer interaction large model.

A training data store may contain training data associated with monitoring node values during normal operation of an industrial asset and simulated abnormal data. An offline model tuning platform accesses the training data from normal operation of the industrial asset and the simulated abnormal data in the training data store. Based on the training data from normal operation of the industrial asset, the simulated abnormal data, an abnormal operating condition, and a constrained optimization solution, controller tuning parameters are created for at least one tuned data-driven adaptive controller such that an operating condition of the industrial asset will move from the abnormal operating condition to a normal operation condition through a stable trajectory. An online monitoring platform receives a stream of current monitoring node values and, when the abnormal operating condition is detected, utilizes the controller tuning parameters to implement the at least one tuned data-driven adaptive controller.

The application relates to the technical field of data analysis, and discloses a data analysis method based on model tuning, which comprises the following steps: when a model tuning control instruction is received, selecting a plurality of groups of hyperparameter combinations from hyperparameter combinations corresponding to preset type data according to a mapping relationship between data types and hyperparameter combinations; determining a target hyperparameter combination from the plurality of groups of hyperparameter combinations based on a first algorithm; constructing an initial data analysis model based on the target hyperparameter combination; performing first model variable equivalent simplification on the initial data analysis model based on a second algorithm to generate an intermediate data analysis model; performing second model variable equivalent simplification on the intermediate data analysis model based on a third algorithm to generate a target data analysis model; and calling the target data analysis model to perform data analysis on preset type data to obtain a data analysis result. The application can effectively improve the running performance of a data analysis model and the accuracy of data analysis.

This invention discloses a method for debugging a lubricating oil system based on digital twins, comprising the following steps: 1) Constructing a simulation model of the lubricating oil system and equipment by defining a lubricating oil system equipment model library; 2) The online model receives real-time operating data of the physical lubricating oil system collected by sensors, transmits the online measured data to the simulation model to drive simulation calculation, and transmits the calculation results to the demonstration model for simulation animation display; 3) Based on the lubricating oil user flow requirements under different pre-given operating conditions, and based on virtual-real fusion simulation, the action parameters of each actuator of the lubricating oil system are calculated through offline model tuning and transmitted to the control system as virtual debugging signals; 4) The physical lubricating oil system receives the control signals from the control system to adjust and control the operation of the physical lubricating oil system. This invention, by constructing a high-precision lubricating oil system model and using digital twin-based virtual debugging technology for lubricating oil systems, can be used for efficient debugging, accurate analysis, and virtual verification of physical systems.

The invention relates to the technical field of process engine control, and discloses a training planning process control method and system based on an Activity process engine. The method comprises the following steps: extracting an instance state snapshot from a process engine, analyzing to construct a time trajectory map of process evolution, marking a key path and a branch point, and generating a process evolution mode template library; and the current evolution stage is accurately determined by matching the real-time process with the template. And then dynamically loading a corresponding control strategy, analyzing the rule to generate a node-level operation instruction, and converting the node-level operation instruction into a model adjustment command which can be identified by an engine. Synchronously monitoring the synchronization state of the resource consumption and the flow track, and automatically calculating a resource quota adjustment scheme when deviation occurs. And finally, cooperatively scheduling a model adjustment command and a resource scheme, forming a control instruction set, and distributing the control instruction set to an engine and a resource manager to realize efficient and adaptive evolution of a flow according to an expected path. According to the invention, the intelligent level of process management and the resource utilization efficiency are improved.

The invention relates to a model self-adaptive optimization method and device, electronic equipment and a readable medium, and the method comprises the steps: obtaining a prediction model corresponding to target equipment, and enabling the prediction model to be obtained through training according to the historical fault data of the target equipment; in the process of performing fault prediction on the target equipment by using the prediction model, monitoring a prediction error of the prediction model and an operation condition of the target equipment; judging whether to trigger a model adjustment mechanism according to the prediction error and the operation condition; under the condition that a model adjustment mechanism is triggered, a corresponding target adjustment strategy is selected according to the size of the prediction error and the change of the operation condition, and the target adjustment strategy is a parameter adjustment strategy or a structure adjustment strategy; and adjusting the prediction model according to the target adjustment strategy to complete adaptive optimization of the prediction model. It is ensured that the model can continuously adapt to equipment state changes, and accurate prediction of equipment faults is achieved.

In a described embodiment, a system for predictive modeling is provided. The system includes a data input module configured to acquire raw data and a data pre-processing module 5 configured to process the raw data to generate sample data. The system further includes a feature engineering module configured to generate derived features using one or more specified criterion and a model generation and optimization module configured to generate and refine a predictive model using a plurality of machine learning techniques. A model evaluation module is configured to assess model performance using predefined metrics. A 10 model tuning system is configured to iteratively adjust model parameters of the predictive model based on performance feedback and a model deployment module is configured to implement the predictive model in a production environment.

A device and method with model determination are provided. The method includes identifying a pre-trained first model and at least one second model tuned based on the first model, for each of layers included in the first model, identifying a first weight value of the first model and at least one second weight value of the at least one second model, for each of the layers, determining a third weight value based on a location derived through a linear interpolation using a first location corresponding to the first weight value and at least one second location corresponding to the at least one second weight value that are in a weight value space, and determining a target model based on the first model including the third weight value.

The invention provides a model adjustment method and electronic equipment, and is applied to the technical field of artificial intelligence. The model adjusting method comprises the steps that first evaluation values of a plurality of model parameters are determined according to calibration data, and the first evaluation values represent the importance degree of the model parameters to a model output result; dividing the plurality of model parameters into a plurality of groups; and determining the calculation precision of the model parameters in the group according to the first evaluation values of the plurality of model parameters in the group, so that the overall calculation precision of the model is smaller than the target calculation precision.

Methods, systems, and devices for wireless communications are described. A user equipment (UE) may obtain data samples for a first machine learning model associated with a task at the UE. The first set of parameters may be associated with the first machine learning model. The UE may transmit a capability message indicating a capability of the UE to perform a tuning procedure of the first machine learning model, and the UE may perform the tuning procedure of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the capability of the UE to perform the tuning procedure of the first machine learning model. The capability of the UE to perform the tuning procedure may be one of an online tuning capability or an offline tuning capability.

The present application discloses a method, system, and computer system for building a model associated with a dataset. The method includes receiving a data set, the dataset comprising a plurality of keys and a plurality of key- value relationships, determining a plurality of models to build based at least in part on the dataset, wherein determining the plurality of models to build comprises using the dataset format information to identify the plurality of models, building the plurality of models, and optimizing at least one of the plurality of models.

The invention discloses a model fine tuning method and device, a storage medium, equipment and a program product, and the method comprises the steps: obtaining the global gradient information of a pre-trained basic model on general data, and enabling the global gradient information to represent the general capability of the basic model; when the domain data is used for fine tuning, a corrected optimization target is generated according to the correlation degree between the domain gradient in the current training step and the global gradient information, and the optimization target is configured to be capable of enhancing domain knowledge learning and inhibiting the universal capability; and the model parameters of the basic model are updated based on the optimization target to generate the target model with the universal capability and the domain knowledge, so that universal learning and domain learning can be decoupled, adaptive learning of the domain performance is realized, the cost is effectively reduced, the model adjustment efficiency is improved, and the domain performance and universal performance are effectively balanced.

This application discloses a data processing method performed by a computer device. The method includes: transmitting K images photographed of an object to a server, where the server obtains K first prediction results by using an image recognition model; constructing a fine-tuning training set according to the K images and the K first prediction results; obtaining a second prediction result of each image in the fine-tuning training set by using a to-be-trained model; updating a model parameter of the to-be-trained model according to the second prediction result of each image and the first prediction result of the image in the fine-tuning training set, to obtain a local recognition model and a model adjustment parameter; and transmitting the model adjustment parameter to the server if a model fine-tuning condition is satisfied, so that the server updates a model parameter of the image recognition model according to a model adjustment parameter set.

The invention is suitable for the technical field of metal material prediction, and provides a high-entropy alloy yield strength prediction method in combination with SSAKOA and based on an ETR-GBR model, and the method comprises the steps: obtaining the original data set information of the high-entropy alloy yield strength, carrying out the preprocessing of the original data set information of the high-entropy alloy yield strength, and obtaining the high-entropy alloy yield strength. Inputting the optimal feature subset information, the hyper-parameter and the fusion weight information into the ERT-GBR model on the basis of the optimal feature subset information, the hyper-parameter and the fusion weight information, and training an extreme random tree model and a gradient boosting regression model on the basis of training set information to generate a trained ERT-GBR model. According to the method, feature screening and model adjustment participation fusion weight determination can be completed in a single optimization framework, the calculation cost brought by repeated grid / random search is remarkably reduced, the convergence speed and stability are improved, the yield strength prediction precision and robustness are improved by relying on the complementarity of ETR and GBR and the adaptive search mechanism of SSAKOA, and the prediction accuracy and robustness of the yield strength are improved. And rapid prediction of the yield strength of the high-entropy alloy is realized.

The invention relates to a model tuning method and device based on prompt, electronic equipment and a readable medium, and the method comprises the steps: obtaining a test data set; calling a prompt optimization model obtained through training in advance, and optimizing the first prompt of the target model by using the test data set and the prompt optimization model to obtain a second prompt; calling a parameter optimization model obtained through training in advance, and adjusting model parameters of the target model by using the test data set, the second prompt and the parameter optimization model to obtain an adjusted target model; the adjusting and optimizing effect of the adjusted target model is detected, when the adjusting and optimizing effect reaches the expected effect, an adjusting and optimizing strategy of the target model is generated, and the adjusting and optimizing strategy comprises a prompt optimizing strategy obtained through the prompt optimizing model and a parameter adjusting strategy obtained through the parameter optimizing model. The problem that the optimal tuning strategy is difficult to find quickly by manually tuning the model during model switching is solved.

1. Name of the product in this design: Graphical User Interface for Bit-Large Model Platform Model Tuning for Electronic Devices. 2. Purpose of this design: For displaying programs and interactions. 3. The key design features of this product are: the graphical user interface content that displays information. 4. The image or photo that best illustrates the design points: Interface change state diagram 2. 5. Since the design points have been clearly expressed, the rear view, left view, right view, top view, and bottom view are omitted. 6. Purpose of the graphical user interface: This graphical user interface is used for model tuning on the bit-large model platform. The main view is the login interface of the BitBigModel platform. In the main view, after the user enters their account and password, they click the "Login" button to enter the interface change state diagram 1, which displays the BitBig Model Platform model management interface. In the interface change state diagram 1, the user clicks the "Model Tuning" column on the left, and then clicks the "Supervised Fine-Tuning (SFT)" item in the drop-down box to enter the interface change state diagram 2, which displays the supervised fine-tuning interface. In interface change state diagram 2, the user clicks the "Create Training Task" button and enters interface change state diagram 3, which displays the Create Training Task interface.

The embodiment of the invention discloses a method and device based on a Fine-tuning screening model tuning method, equipment and a storage medium. The method comprises the steps of obtaining data features of a to-be-optimized model; according to the data features and meta-information of each Fine-tuning method in a pre-constructed Fine-tuning method library, determining a matching index of each Fine-tuning method and the model to be adjusted and optimized; and selecting a target Fine-tuning method for the model to be adjusted and optimized according to the matching index. On the basis, the target Fine-tuning method suitable for the data features can be found through the data features of the model to be adjusted and optimized, so that the model to be adjusted and optimized is more adapted during subsequent adjustment and optimization, and the model adjustment and optimization effect is improved.

The invention belongs to the field of time sequence pre-training large model characteristic testing, and provides a time sequence pre-training large model key characteristic rapid testing method and a related device, and the method comprises the steps: constructing a test set, inputting the test set into a to-be-tested time sequence pre-training model, and determining and outputting a model with basic and composite time pattern recognition capability; sequentially adding different types of external features into the original target sequence, respectively calculating baseline prediction performance of only using the original target sequence and prediction performance after adding each feature, and outputting a feature redundancy evaluation result; and based on a feature redundancy evaluation result, observing a performance curve of prediction performance of the model changing along with the context length, and based on the performance curve, judging an optimal memory window of the model and sensitivity of the optimal memory window to the long context. The test process has rapidness and systematicness, the basic portrait of the model can be obtained through rapid diagnosis for several minutes, and direction guidance is provided for deep development of model tuning.

The invention discloses a motion pattern recognition model training method, device and system fusing misjudgment data, and belongs to the field of model training. The model training method comprises the steps that an initial recognition model and a sample data set are acquired, a modulation coefficient and a proportionality coefficient are introduced when a loss function of the initial recognition model is defined, a weight factor does not need to be independently set for each category, the number of hyper-parameters is small, parameter adjustment cost and structural complexity of the model are reduced, and training efficiency is improved. The model training practice is shortened and the training cost is reduced while the recognition accuracy of the exoskeleton motion mode recognition model is ensured. Meanwhile, double labeling is carried out, the misclassification feature utilization rate is increased, and the real-time performance and accuracy of a motion mode recognition result are enhanced.

Disclosed herein are a local language model tuning apparatus and method. The local language model tuning apparatus is configured to align a local language model using a first split subset of a first dataset implemented as a list of pairs of a prompt and a response of a service language model, perform batch inference of obtaining a result sample by inputting a prompt recorded in a second split subset of the first dataset to the aligned local language model, evaluate performance of the aligned local language model through the service language model based on the result sample, and when an evaluation score obtained by evaluating the performance of the aligned local language model exceeds a preset threshold, deploy the aligned local language model.