15 results about "Fixed sequence" patented technology

The invention discloses an intelligent production scheduling method and system based on a fixed sequence genetic algorithm. The method comprises the following steps: dividing independent production areas of a workshop; key operation is determined, and modeling is carried out on time of the key operation; generating a process set containing a fixed sequence key job based on a process path, performing crossover variation by using an improved genetic algorithm, and keeping the sequence of internal processes of a task unchanged to avoid an invalid solution; and calculating the shortest processing duration through population iteration, and generating a production scheduling Gantt chart. The system comprises a workshop function area module, a key operation type module, a key operation modeling module and a production scheduling result output module, so as to realize the output of an intelligent production scheduling result. The method improves the production scheduling efficiency and the applicability of the production scheduling result, and is suitable for complex production scenes such as electronic assembly and cabinet debugging.

The application discloses a chip silicon pre-verification method and device, a hardware simulator and a medium, and belongs to the technical field of artificial intelligence chips. The method comprises the following steps: acquiring memory data from a physical accelerator; executing initial logic of a plurality of operators to be verified in a fixed order; testing the plurality of operators in sequence to obtain standard output results corresponding to the plurality of operators respectively; and storing the standard output results of each operator in the plurality of operators in target memories corresponding to the operators. When each verification operator is verified individually, the standard output results of the previous operator are directly acquired from the target memory corresponding to the previous operator as simulation input data of the verification operator, the reconstruction logic of the verification operator is executed, the first simulation output result of the verification operator is obtained, and the first verification result of the verification operator is determined. Each operator in an end-to-end system is disassembled, coupling between the operators is avoided, each operator does not need to be analyzed, the operator that fails to verify is accurately positioned, and the verification efficiency is improved.

The invention relates to the technical field of operation and maintenance management, and discloses an operation and maintenance task-oriented automatic template management system, which comprises the following steps of: uniformly expressing an equipment tree in a fixed sequence of projects, branch companies, machine room sub-parts / sub-items and first-level / second-level / third-level equipment through node indexing-hierarchical coding-alignment to generate a stable index vector; cross-project naming differences and hierarchy missing or redundancy are converted into recoverable index offset, the same object is positioned consistently in different tree versions, candidate screening is not affected by hierarchy jitter, a type extraction module executes alias merging on a type table to obtain type mapping, then type vectors are extracted from a task list according to the type mapping, and the type vectors are extracted from the task list. According to the method, alias crossing and granularity inconsistency are converged to a unified key space, ambiguity is eliminated in an extraction link, and template matching is accurately restrained, so that near-synonym error locking is reduced, the accuracy and stability of template automatic locking are improved, and manual proofreading and backspacing are reduced.

The present application relates to the technical field of building piling, in particular to a piling auxiliary equipment for building construction, which comprises a fixed base, a lifting disc, a casing control assembly and a reinforcement cage control assembly, the reinforcement cage control assembly comprises a fixed sleeve, a first movable ring, a control disc, a second movable ring and a hooking plate, a support rod is fixedly arranged at the bottom of the fixed sleeve, the casing control assembly comprises a prismatic column, a protruding plate is fixedly arranged on the first movable ring, a circular through hole is formed in the protruding plate, the control disc is rotatably arranged above the protruding plate, a prismatic hole is formed in the control disc, and the second movable ring is rotatably arranged outside the control disc. The present application reduces the time required for installation, and also makes the entire use process not need to be carried out according to the fixed order, which facilitates the use of the entire equipment and ensures the stability of the reinforcement cage when the casing is shaken, thereby ensuring that the subsequent work will not be affected.

The application discloses an individual risk claim maximum payment implementation method, which comprises the following steps: 1) responsibility matching; 2) amount calculation; 3) sequence determination; 4) amount allocation; 5) circulation settlement; and 6) result verification. Through the dynamic sorting logic of 'the minimum payment amount is given priority to', the application reduces the early deduction loss of high-value responsibility, ensures the maximization of the total payment amount after the superposition of multiple responsibilities, solves the problem that the traditional fixed sequence settlement cannot realize optimal payment, and improves user satisfaction. Through fine processing of the sharing and deduction rules between responsibilities, the payment amount is split according to the payment type, so that the deduction logic is accurate; through the layer-by-layer promotion of multiple settlement formulas, segmented payment rules, limit and premium accounting, the amount calculation is accurate and error-free, and the risk of claim dispute is reduced.

The invention relates to the technical field of intelligent manufacturing, in particular to a vehicle repair store operation system and method using an intelligent station, and the system comprises a scheduling data processing module, a delay evaluation module, a scheduling adjustment module, a priority decision module, and a scheduling execution module. According to the method, quantitative evaluation and grading of the station delay state are achieved by calculating the handover time difference between the operations and comparing the handover time difference with the standard period, the chain reaction range possibly caused by single task delay can be recognized in a prospective mode, and then the adjustment requirement is focused on the affected follow-up tasks; unnecessary global disturbance on an overall plan is avoided, meanwhile, the emergency degrees of the tasks and the time proximity after adjustment are comprehensively evaluated, dynamic station time slot occupation weights are generated, all the tasks to be adjusted are globally sorted, the limitation of a traditional fixed sequence is broken through, and the task scheduling efficiency is improved. And finally, according to the dynamic priority of the task and the actual operation duration, automatically searching and matching an optimal idle station time slot.

A computer-implemented system and method for producing a multi-scene digital video composition through sequential user contributions is disclosed. A processing system enforces structured scene creation by assigning each scene to a defined position within a predetermined sequence and restricting continuation to a single designated contributor or a publicly accessible pool within a defined time constraint. Each submitted scene is validated and permanently locked from modification upon acceptance. The system automatically controls progression between contributors, enforces deadlines, and prevents unauthorized editing or reordering of scenes. Accepted scenes are assembled by the system into a final video composition in a fixed sequence without user intervention. If continuation conditions are not satisfied within the defined time, the system automatically finalizes or suspends the composition based on predefined criteria and publishes the resulting content to a feed.

The application discloses a stacked integrated intelligent control chip with instant action generation script, and belongs to the technical field of intelligent control hardware and heterogeneous packaging technology.The chip adopts a 3D vertical stacked packaging structure, integrates a SIM identity communication chip on the top of a main chip, and realizes high-speed direct connection through a TSV (Through Silicon Via).An instant action generation script is stored in a nonvolatile storage unit inside the main chip, and the script is used for real-time calculation and dynamic generation of continuous action instructions according to external sensing signals, instead of executing a fixed sequence.An action logic operation core and an on-chip cache are built in the chip, and the whole action generation process is completed on the chip without the participation of external memory, CPU or cloud.The application realizes an end-side intelligent hardware architecture with self-determination, instant action generation and identity security after power-on, solves the problems of traditional controllers, such as dependence on external programs, high response delay, weak security and large memory occupation, and is suitable for scenes such as intelligent terminals and robots.

The embodiment of the invention relates to the technical field of artificial intelligence, in particular to a KV Cache data multiplexing method and device and electronic equipment, and the method can improve the efficiency of model prefill. The method comprises the steps of generating a prompt corresponding to the round of questions and answers in response to questions input by a user in the round of questions and answers; the prompt comprises a system prompt part, a user prompt part and a user input part which are arranged according to a fixed sequence; the position ID of each part in the prompt is fixed; the user prompting part comprises a plurality of segmented blocks, and the position ID of each of the plurality of blocks in the prompt is fixed; a user input section including a question; inputting a prompt to the large language model, and querying whether a current block in the plurality of blocks hits the KV Cache data in the memory or not in a Prefill stage; and under the condition of hit, reading KV Cache data corresponding to the current block from the memory.

The invention provides a data receiving method, device and system, and belongs to the technical field of communication. And after the transmission of the data stream sent by the second module is interrupted and the transmission is recovered, the first module can search the fixed sequence in the target range in the data stream after the transmission is recovered. Wherein the target range is determined based on a reference position of a fixed sequence in the data stream, and the reference position is determined based on a fixed time length. As the fixed duration corresponds to the fixed interval between the two adjacent fixed sequences in the data stream, the deviation between the reference position determined based on the fixed duration and the actual position of the fixed sequence is small in the scene that the transmission of the data stream is interrupted and recovered. Correspondingly, in the target range determined based on the reference position, the first module can quickly search the fixed sequence. Namely, according to the scheme provided by the invention, the data volume required to be searched by the first module when the fixed sequence is searched can be reduced, so that the search efficiency of the fixed sequence is improved.

A system for transferring semiconductor workpieces from a load lock to an alignment station and onto a platen is disclosed. The system comprises two load locks, two robots, and one alignment station. Each robot is associated with its respective load lock and follows a fixed sequence. The robot returns the processed workpiece to the load lock and removes the unprocessed workpiece. The robot then moves to the alignment station, where it removes the aligned workpiece from the alignment station and places the unprocessed workpiece onto the alignment station. Next, the robot moves to the platen, where it removes the processed workpiece and places the aligned workpiece. The robot then returns to the load lock and repeats this sequence.

This invention discloses a low-instruction-overhead, reconfigurable normalization operator implementation method based on SIMD vector processing, relating to deep learning algorithm model technology and electronic information technology. The invention first divides the normalization operator into reduction and scalar operations, compiles the operations into a fixed sequence of M microcodes, preloads them into a microcode sequence memory, and configures a SIMD vector processing unit consisting of an instruction decoding unit, a hardware autonomous execution engine, and a memory access address generator. It continuously performs sequence length token and channel group scanning, performing P×N iterations on P groups of channels and N tokens, physically reusing the microcode sequence. The microcode pointer is automatically reset to zero, and intermediate result fusion is adaptively completed by scalar microcode operator encoding. Compared with existing GPU solutions, this method has better reconfigurability, lower execution cycle, and lower instruction overhead.

The invention discloses a chip pre-silicon verification method and device, a hardware simulator and a medium, and belongs to the technical field of artificial intelligence chips.The method comprises the steps that memory data are obtained from a physical accelerator, initial logics of multiple operators to be verified are executed according to a fixed sequence, the multiple operators are tested in sequence, and the tested operators are obtained; the method comprises the following steps: obtaining standard output results corresponding to a plurality of operators, and storing the standard output result of each operator in the plurality of operators in a target memory corresponding to the operator; the method comprises the following steps: directly acquiring a standard output result from a target memory corresponding to a previous operator as analog input data of a verification operator, executing reconstruction logic of the verification operator to obtain a first analog output result of the verification operator so as to determine a first verification result of the verification operator, disassembling each operator in an end-to-end mode, and obtaining a second verification result of the verification operator; it is guaranteed that the operators are not coupled, the operators do not need to be analyzed one by one, the operators failed in verification are accurately positioned, and verification efficiency is improved.

The embodiment of the invention provides a GPU test method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: employing a to-be-tested GPU, carrying out the preset number of times of iteration training of a preset large model through the training data of a fixed sequence, obtaining the model loss of the preset large model in each iteration, and taking the model loss as the target model loss; obtaining a difference characterization value between a sequence formed by the loss of the target model and a reference sequence; wherein the reference sequence comprises model loss obtained by carrying out preset times of round iterative training on a preset large model by using a reference GPU and training data in a fixed sequence; and generating a test result of the to-be-tested GPU based on the difference representation value. According to the embodiment of the invention, the GPU can be tested, so that technicians can know the training effect of the GPU conveniently, and then a proper GPU is selected.

This invention provides a spoke plate, a layup method, and a layup optimization method, belonging to the field of spoke plate layup technology. The spoke plate is a rotating structure with a central axial hole. The rotating cross-section of the spoke plate includes: Block I to Block V, where Block I is the innermost region of the spoke plate, with a constant thickness along the first radial direction; and Block II, connected to Block I, with a constant thickness along the first radial direction. The spoke plate of this invention divides the rotating cross-section into multiple blocks. By laying up each block in a fixed sequence, a superior spoke plate structure is obtained. This invention overcomes the limitations of traditional layup methods by dividing the spoke plate structure into thickness zones, determining the layup method for the carbon fiber spoke plate, and achieving flexibility and intelligence in carbon fiber spoke plate layup optimization by combining CAE analysis with genetic algorithms, resulting in a theoretically optimal layup structure in terms of mechanical properties.