6 results about "Transformation operator" patented technology

The embodiment of the invention provides a multi-source heterogeneous data integration method and device fusing a large model conversion operator. The method comprises the steps that heterogeneous data are collected from a multi-source heterogeneous data source, preprocessing operation of data cleaning is carried out, and preprocessed data features are obtained; inputting the preprocessed data features and the target format into a large language model, and generating a conversion operator including data structure analysis, field mapping and type adaptation; and based on the conversion operator, performing distributed parallel data conversion and integration in a distributed computing framework, and integrating and verifying the integrated data. According to the scheme, by introducing the intelligent reasoning ability of a large language model, the efficiency optimization of distributed calculation and the quality verification mechanism of the whole process, the core problems of the traditional data integration technology in the aspects of rule stiffness, high manual dependency and quality control deficiency are systematically solved.

The embodiment of the present application provides an optimization method, device, storage medium and program product for attention mechanism calculation, which relates to the field of artificial intelligence technology. The method includes: executing a forward recalculation operator based on the query block and the key block to obtain a forward normalization result. Executing a first matrix multiplication operator based on the output gradient block and the forward normalization result to obtain a value intermediate result, and simultaneously executing a dot product summation operator based on the forward output block and the output gradient block in parallel to obtain an intermediate tensor. Executing a first type of transformation operator on the value intermediate result to obtain a value gradient block; and simultaneously executing a gradient backpropagation operator based on the output gradient block, the value block, the intermediate tensor and the forward normalization result in parallel to obtain a reverse normalization result. Finally, based on the reverse normalization result, the key block and the query block, the query gradient block and the key gradient block are obtained, thereby achieving effective pipeline processing of different operators, thereby improving the overall performance of the attention mechanism calculation.

The invention discloses a tensor arrangement change compiling method, a tensor arrangement change compiling device and a computer readable storage medium. The compiling method is applied to an artificial intelligence model, and the compiling method comprises the following steps: obtaining a computational graph of the artificial intelligence model; carrying out dynamic analysis on the calculation graph, identifying the existing tensor arrangement conversion operation and the position where tensor arrangement conversion needs to be inserted, and inserting an arrangement conversion operator at the corresponding position so as to update the calculation graph; analyzing the updated calculation graph to obtain an analysis result; simplifying redundant tensor operation according to an analysis result, and forming a corresponding operator demand; an optimization operator is generated by combining operator requirements and hardware design; and forming a compilation instruction for the computational graph according to the optimization operator. According to the technical scheme, the tensor conversion efficiency can be effectively improved.

The application provides a method for inserting a conversion operator, which comprises: determining an initial data format of a first operator in a neural network; determining a current data format of the first operator; when the initial data format of the first operator is different from the current data format, inserting at least one first conversion operator and at least one second conversion operator around the first operator, the first conversion operator being used for converting input data of the first operator from the initial data format to the current data format, and the second conversion operator being used for converting output data of the first operator from the current data format to the initial data format. In the method, the initial data format is a data format beneficial to hardware performance, and the current data format is a data format beneficial to improving processing efficiency of the first operator; therefore, the conversion operator is inserted according to the initial data format and the current data format of the first operator, so that the requirements of the operator and the requirements of hardware can be met simultaneously.

The invention provides a large language model operator conversion system and method and an AI accelerator, the system comprises an operator identification module, a conversion module and a model reconstruction module, the operator identification module analyzes a computational graph structure of an input model to identify convertible operators meeting preset conversion conditions in the computational graph structure; the conversion module converts the convertible operator into an equivalent operator sequence based on a preset conversion strategy; and the model reconstruction module replaces nodes corresponding to convertible operators in the computational graph of the input model with an equivalent operator sequence, updates the connection relation and the weight initializer, and generates an updated model. According to the system, pipeline processing of model optimization is realized through an identification rule and a conversion strategy, and adaptation for different hardware platforms can be realized by updating a preset conversion condition and a conversion strategy, so that the problems of low deployment efficiency and low expandability on a CNN (Convolutional Neural Network) accelerator are solved.

The application discloses a model design method and device, equipment and a storage medium, and relates to the technical field of computers, which comprises the following steps: obtaining an original model, and obtaining an operator satisfying a preset condition from the original model to obtain a target operator; inserting a first conversion operator and a second conversion operator in front of the target operator; distributing the target operator to a target compiler, and distributing the first conversion operator and the second conversion operator to a central processing unit to obtain a compiled model. The application increases the conversion operator in front of the operator satisfying the preset condition, and distributes the conversion operator to the central processing unit, so that the central processing unit converts the data that cannot be used by the target operator into usable data, and then sends the data to the target compiler for processing, thereby realizing the conversion of the data mode, improving the equipment utilization rate of the target compiler and the processing speed of the model.