Adaptive rendering decision-making system and method based on AI large model, storage medium and equipment

Abstract

The invention discloses an adaptive rendering decision-making system and method based on an AI large model, a storage medium and equipment, and the system comprises the steps: a rendering decision-making manager is internally provided with a rendering decision-making library which is used for storing rendering decisions, requesting an experiment rendering decision from an experiment manager according to the type of to-be-rendered equipment, and issuing the rendering decision; an experiment pool is arranged in the experiment manager and is used for storing an experiment rendering decision and scheduling the experiment generator to create an experiment of the to-be-rendered equipment; the rendering configuration generator generates a plurality of experiment rendering decisions according to the created experiment, stores the experiment rendering decisions in an experiment pool, and sequentially issues the experiment rendering decisions to the to-be-rendered equipment for rendering to generate experiment rendering feedback; and the rendering decision scoring device generates experimental rendering decision scores according to the experimental rendering feedback, sorts the experimental rendering decisions according to the sequence of the experimental rendering decision scores from high to low, and stores the first m experimental rendering decisions in a rendering decision library as rendering decisions. According to the method, the rendering decision accuracy is improved, and manual dependence is avoided.

Description

Technical Field

[0001] The present invention relates to the field of graphics and image rendering technology, and in particular to an adaptive rendering decision system, method, storage medium, and device based on an AI large model. Background Art

[0002] Graphics and image rendering is the core and foundation of video presentation. The choice of rendering method and configuration settings are crucial to the quality of video presentation performance. A common approach currently involves developers creating a whitelist based on their understanding of the device, specifying which device models correspond to which configurations, and also setting a universal default rendering configuration. Another approach is to make pre-rendering selections based on the device's rendering requirements, such as supported GPU types and vendors. This selection is often based on experience.

[0003] The advantages of these two configuration methods are that they are relatively simple, direct, and easy to implement, but their disadvantages are also obvious: the first method relies too much on the developer's experience. If the developer lacks experience, there will not be a good whitelist, and the default decoding configuration used by devices outside the whitelist often has poor rendering effects. This problem will become more serious as the number of models that need to be supported increases. The other method uses static judgment. The same judgment results may vary greatly on different models. For example, although many Android models support Vulkan rendering, the actual rendering efficiency is low, even far inferior to OpenGL; or although the hardware determines that Vulkan rendering is supported, the actual rendering will fail or have other problems such as abnormal effects. Therefore, the second method is often unreliable for Android devices with severe model fragmentation. Summary of the Invention

[0004] In response to the problems existing in the prior art, the present invention provides an adaptive rendering decision system, method, storage medium and device based on an AI big model. The AI big model is used to intelligently score each experimental rendering decision to obtain the optimal rendering decision suitable for the device model, thereby improving the accuracy of the rendering decision.

[0005] To achieve the above technical objectives, the present invention adopts the following technical solutions: an adaptive rendering decision system based on an AI large model, comprising: a rendering decision manager, an experiment manager, an experiment generator, a rendering configuration generator, and a rendering decision scorer;

[0006] The rendering decision manager is provided with a rendering decision library for storing rendering decisions; the rendering decision manager also requests the experimental rendering decision from the experiment manager according to the type of the device to be rendered, and issues the rendering decision;

[0007] The experiment manager is provided with an experiment pool for storing experimental rendering decisions; the experiment manager also schedules the experiment generator to create experiments for devices to be rendered;

[0008] The rendering configuration generator generates several experimental rendering decisions based on the created experiment, saves them in the experiment pool, and sends them to the devices to be rendered in sequence for rendering, generating experimental rendering feedback;

[0009] The rendering decision scorer generates an experimental rendering decision score according to the experimental rendering feedback, sorts the experimental rendering decisions in descending order of the experimental rendering decision score, and saves the first m experimental rendering decisions as rendering decisions into a rendering decision library.

[0010] Furthermore, the rendering decision is the same as the experimental rendering decision, including: interface type, rendering mechanism, whether to enable multi-GPU configuration, resolution, and whether to support ray tracing.

[0011] Furthermore, the experimental rendering feedback includes: whether the renderer is initialized normally and whether the rendering is successful; if the rendering is successful, the rendering duration is fed back; if the rendering fails, whether the rendering is a black screen or a green screen is fed back.

[0012] Furthermore, if the renderer fails to initialize normally or fails to render, the rendering decision scorer directly records the experimental rendering decision score as 0; otherwise, the experimental rendering decision score is generated by the AI big model.

[0013] Furthermore, the process of generating the experimental rendering decision score through the AI big model is: generating the context of the AI big model based on the accumulated experimental rendering feedback of the model of the device to be rendered, the corresponding experimental rendering decision and the model information of the device to be rendered, and using the Prompt template of the AI big model to generate a Prompt instance for the current experimental rendering decision and experimental rendering feedback, and transmitting it to the AI big model to generate the experimental rendering decision score.

[0014] Furthermore, the Prompt template is: for the model of the device to be rendered {}, when the interface type is {}, the rendering mechanism is {}, the multi-GPU configuration is {}, the resolution is {}, the ray tracing support is {}, and the {}th experimental rendering feedback is {}, please score the current experimental rendering decision, and the score range is between 0 and 100.

[0015] Furthermore, the present invention also provides an adaptive rendering decision method based on an AI large model, which uses the adaptive rendering decision system based on an AI large model and specifically includes the following steps:

[0016] Step S1: Retrieve the rendering decision corresponding to the model of the device to be rendered from the rendering decision library of the rendering decision manager. If the retrieval is successful, directly send the rendering decision with the highest rendering decision score to the device to be rendered for rendering; otherwise, request the rendering decision from the experiment manager;

[0017] Step S2: The experiment manager schedules the experiment generator to create an experiment for the device to be rendered. The rendering configuration generator generates several experimental rendering decisions based on the created experiment, saves them to the experiment pool, and sends them to the device to be rendered in sequence for rendering, generating experimental rendering feedback.

[0018] Step S3: The rendering decision scorer generates an experimental rendering decision score based on the experimental rendering feedback, sorts the experimental rendering decisions in descending order of the experimental rendering decision score, and saves the first m experimental rendering decisions as rendering decisions into a rendering decision library;

[0019] Step S4: The rendering decision management server sends the rendering decision with the highest rendering decision score to the device to be rendered for rendering.

[0020] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the adaptive rendering decision method based on the AI large model.

[0021] Furthermore, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the adaptive rendering decision method based on the AI large model is implemented.

[0022] Compared with the existing technology, the present invention has the following beneficial effects: the adaptive rendering decision system and method based on the AI big model of the present invention can create experiments for the device to be rendered, and generate different experimental rendering decisions and send them to the device to be rendered for rendering, and generate experimental rendering feedback. By introducing a rendering decision scorer, the experimental rendering feedback is intelligently scored based on the AI big model, and the big data intelligent capabilities of the AI big model are utilized to continuously accumulate and learn the rendering feedback of the same model device, and intelligently score each experimental rendering decision under the model, and use this as the basis for generating the optimal rendering decision for the model, which greatly improves the accuracy and acquisition efficiency of the optimal rendering decision, avoids the dependence on developer experience in obtaining the optimal rendering decision, and at the same time, effectively avoids the rendering inefficiency problem that may be caused by the statically specified decoding method, and realizes a more comprehensive rendering decision selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of the adaptive rendering decision system based on the AI big model of the present invention;

[0024] Figure 2 A schematic diagram of the scoring performed by the rendering decision scorer in the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be further explained below with reference to the accompanying drawings.

[0026] like Figure 1 This is a schematic diagram of the adaptive rendering decision system based on the AI big model of the present invention. The adaptive rendering decision system includes: a rendering decision manager, an experiment manager, an experiment generator, a rendering configuration generator and a rendering decision scorer.

[0027] The rendering decision manager is provided with a rendering decision library for storing rendering decisions; the rendering decision manager also requests experimental rendering decisions from the experiment manager based on the type of device to be rendered, and issues rendering decisions for rendering of the device to be rendered; the rendering decision in the present invention has the same content as the experimental rendering decision, and both are a set of parameters required to build a renderer, including: interface type, rendering mechanism, whether to open multi-GPU configuration, resolution and whether to support ray tracing.

[0028] The experiment manager has an experiment pool for storing experimental rendering decisions; the experiment manager also schedules the experiment generator to create experiments for the devices to be rendered.

[0029] The rendering configuration generator generates several experimental rendering decisions based on the created experiment, saves them in the experiment pool, and sends them to the rendering device for rendering in sequence, generating experimental rendering feedback, including: whether the renderer is initialized normally and whether the rendering is successful; if the rendering is successful, feedback on the rendering time; if the rendering fails, feedback on whether the rendering is a black screen or a green screen.

[0030] The rendering decision scorer generates an experimental rendering decision score based on the experimental rendering feedback, sorts the experimental rendering decisions in descending order of the experimental rendering decision score, and saves the top m experimental rendering decisions as rendering decisions into the rendering decision library.

[0031] The experimental rendering decision in the present invention includes static experimental rendering decision and dynamic experimental rendering decision, so as to obtain a more comprehensive and accurate rendering decision judgment. Among them, the static experimental rendering decision includes obtaining: CPU main frequency, CPU instruction set, CPU manufacturer, GPU frequency, GPU model, GPU temperature, system type, system version and memory size; the dynamic experimental rendering decision includes obtaining: CPU average usage rate, CPU average temperature, memory average occupancy rate and actual usage rate of different rendering methods. According to the scores of the static experimental rendering decision and the dynamic experimental rendering decision, it is calculated whether the best rendering method for the device model to be rendered is OpenGL rendering or Vulkan rendering. For example, if Android supports Vulkan rendering, it can make more complex and flexible settings, and use OpenGL rendering as the default fallback strategy. If the scores of the static experimental rendering decision and the dynamic experimental rendering decision are both less than the OpenGL experimental score, the OpenGL-based rendering decision is adopted.

[0032] In the present invention, if the renderer fails to initialize normally or fails to render, the rendering decision scorer directly records the experimental rendering decision score as 0; otherwise, the experimental rendering decision score is generated by the AI big model, such as Figure 2 The specific process is as follows: The AI model context is generated based on the accumulated experimental rendering feedback, corresponding experimental rendering decisions, and the model information of the device to be rendered. The current experimental rendering decision and experimental rendering feedback are generated using the AI model's Prompt template to generate a Prompt instance, which is then transmitted to the AI model to generate a score for the experimental rendering decision. Leveraging the AI model's big data intelligence capabilities, the AI model continuously accumulates and learns rendering feedback for the same device model, intelligently scoring each experimental rendering decision for that model. This score serves as the basis for generating the optimal rendering decision for that model, greatly improving the accuracy and efficiency of obtaining the optimal rendering decision and avoiding reliance on developer experience to obtain the optimal rendering decision.

[0033] The AI large model used in the present invention can be a large language model (LLM). The prompt template in the present invention is: for the model of the device to be rendered {}, when the interface type is {}, the rendering mechanism is {}, the multi-GPU configuration is {}, the resolution is {}, the ray tracing support is {}, and the {}th experimental rendering feedback is {}, please score the current experimental rendering decision, and the score range is between 0 and 100.

[0034] In one technical solution of the present invention, an adaptive rendering decision method based on an AI large model is also provided, which specifically includes the following steps:

[0035] Step S1: Retrieve the rendering decision corresponding to the model of the device to be rendered from the rendering decision library of the rendering decision manager. If the retrieval is successful, directly send the rendering decision with the highest rendering decision score to the device to be rendered for rendering; otherwise, request the rendering decision from the experiment manager;

[0036] Step S2: The experiment manager schedules the experiment generator to create an experiment for the device to be rendered. The rendering configuration generator generates several experimental rendering decisions based on the created experiment, saves them to the experiment pool, and sends them to the device to be rendered in sequence for rendering, generating experimental rendering feedback.

[0037] Step S3: The rendering decision scorer generates an experimental rendering decision score based on the experimental rendering feedback, sorts the experimental rendering decisions in descending order of the experimental rendering decision score, and saves the first m experimental rendering decisions as rendering decisions into a rendering decision library;

[0038] Step S4: The rendering decision management server sends the rendering decision with the highest rendering decision score to the device to be rendered for rendering.

[0039] The adaptive rendering decision method based on the AI big model of the present invention can create an experiment for each model to dynamically verify the execution of different experimental rendering decisions, and make full use of the AI big model to intelligently score each experimental rendering decision according to the execution feedback, and sort them according to the experimental rendering decision scores, so as to obtain the best rendering decision plan for the model, and dynamically send it to the corresponding device model, effectively avoiding the rendering inefficiency problem that may be caused by the use of statically specified decoding methods. The optimal rendering decision obtained by the present invention will no longer be limited by the developer's experience, nor will it be based on a simple judgment of the device's capabilities. Instead, it is based on the AI intelligent scoring of the device's actual rendering performance, which is objective; and its self-learning mechanism determines that no matter how many new models appear, the optimal rendering decision can be quickly obtained through experiments. After obtaining the optimal rendering decision, the experiment can be terminated, and its occupied resources can be effectively recovered for other experiments, thereby ensuring that its occupied resources are not too large.

[0040] In one technical solution of the present invention, a computer-readable storage medium is also provided, storing a computer program, which enables a computer to execute the adaptive rendering decision method based on the AI large model.

[0041] In one technical solution of the present invention, an electronic device is also provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the adaptive rendering decision method based on the AI large model is implemented.

[0042] In the embodiments disclosed herein, computer storage media can be tangible media that can contain or store programs for use by or in conjunction with an instruction execution system, device, or apparatus. Computer storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0043] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0044] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. An adaptive rendering decision system based on AI big model, characterized by: include: Rendering Decision Manager, Experiment Manager, Experiment Generator, Rendering Configuration Generator, and Rendering Decision Scorer; The rendering decision manager is provided with a rendering decision library for storing rendering decisions; the rendering decision manager also requests the experimental rendering decision from the experiment manager according to the type of the device to be rendered, and issues the rendering decision; The experiment manager is provided with an experiment pool for storing experimental rendering decisions; the experiment manager also schedules the experiment generator to create experiments for devices to be rendered; The rendering configuration generator generates several experimental rendering decisions based on the created experiment, saves them in the experiment pool, and sends them to the devices to be rendered in sequence for rendering, generating experimental rendering feedback; The rendering decision scorer generates an experimental rendering decision score according to the experimental rendering feedback, sorts the experimental rendering decisions in descending order of the experimental rendering decision score, and saves the first m experimental rendering decisions as rendering decisions into a rendering decision library.

2. The adaptive rendering decision system based on AI big model according to claim 1, characterized in that: The rendering decision described above is the same as the experimental rendering decision, including: interface type, rendering mechanism, whether to enable multi-GPU configuration, resolution, and whether to support ray tracing.

3. The adaptive rendering decision system based on AI large model according to claim 1, characterized in that: The experimental rendering feedback includes: whether the renderer is initialized normally and whether the rendering is successful; if the rendering is successful, the rendering time is fed back; if the rendering fails, the rendering is fed back whether it is a black screen or a green screen.

4. The adaptive rendering decision system based on AI large model according to claim 3 is characterized in that: If the renderer fails to initialize normally or fails to render, the rendering decision scorer directly records the experimental rendering decision score as 0; otherwise, the experimental rendering decision score is generated by the AI big model.

5. The adaptive rendering decision system based on AI large model according to claim 4 is characterized in that: The process of generating the experimental rendering decision score through the AI big model is as follows: generate the context of the AI big model based on the accumulated experimental rendering feedback of the model of the device to be rendered, the corresponding experimental rendering decision and the model information of the device to be rendered, and use the Prompt template of the AI big model to generate a Prompt instance for the current experimental rendering decision and experimental rendering feedback, and transmit it to the AI big model to generate the experimental rendering decision score.

6. The adaptive rendering decision system based on AI large model according to claim 5, characterized in that: The prompt template is: for the model of the device to be rendered {}, when the interface type is {}, the rendering mechanism is {}, the multi-GPU configuration is {}, the resolution is {}, the ray tracing support is {}, and the {}th experimental rendering feedback is {}, please score the current experimental rendering decision, with the score range between 0 and 100.

7. An adaptive rendering decision method based on AI large model, characterized in that: The adaptive rendering decision system based on the AI large model according to any one of claims 1 to 6 specifically comprises the following steps: Step S1: Retrieve the rendering decision corresponding to the model of the device to be rendered from the rendering decision library of the rendering decision manager. If the retrieval is successful, directly send the rendering decision with the highest rendering decision score to the device to be rendered for rendering; otherwise, request the rendering decision from the experiment manager; Step S2: The experiment manager schedules the experiment generator to create an experiment for the device to be rendered. The rendering configuration generator generates several experimental rendering decisions based on the created experiment, saves them to the experiment pool, and sends them to the device to be rendered in sequence for rendering, generating experimental rendering feedback. Step S3: The rendering decision scorer generates an experimental rendering decision score based on the experimental rendering feedback, sorts the experimental rendering decisions in descending order of the experimental rendering decision score, and saves the first m experimental rendering decisions as rendering decisions into a rendering decision library; Step S4: The rendering decision management server sends the rendering decision with the highest rendering decision score to the device to be rendered for rendering.

8. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables the computer to execute the adaptive rendering decision method based on the AI large model as described in claim 7.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the adaptive rendering decision method based on the AI large model as claimed in claim 7 is implemented.

Images