VRB-based model training data high-speed transmission communication method

Abstract

The embodiment of the invention provides a VRB-based data transmission system, method, device and equipment and a medium, the system comprises first equipment and second equipment, and the first equipment and the second equipment are in communication connection based on an articulated naturality web; the first equipment comprises a first model and a first network card, and is used for packaging data used by the first model through the first network card based on an articulated naturality web protocol to obtain a data packet, and sending the data packet to the second equipment based on the articulated naturality web; the second equipment comprises a second model and a second network card and is used for analyzing the data packet through the second network card to obtain data, transmitting the data to the second model and training based on the data through the second model; according to the method, the problem that computing resources are occupied is effectively avoided, the data transmission delay between different equipment models is reduced, and meanwhile, the data transmission efficiency is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a VRB-based data transmission system, method, apparatus, device, and medium. Background Technology

[0002] In recent years, with the development of technology, the rapid iteration of AI technology has driven the emergence of new applications, which has posed many severe challenges to data transmission systems. When transmitting data between models of different devices, the network protocol stack consumes a lot of CPU resources, which crowds out computing resources, leading to increased data transmission latency and reduced transmission efficiency. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention are proposed to provide a VRB-based data transmission system, method, apparatus, device, and medium that overcomes or at least partially solves the above problems.

[0004] To address the aforementioned issues, this invention discloses a VRB-based data transmission system, comprising: a first device and a second device, wherein the first device and the second device are connected via video network communication. The first device includes a first model and a first network interface card (NIC), and is used to encapsulate the data used by the first model based on the video network protocol to obtain a data packet, and send the data packet to the second device based on the video network protocol; The second device includes a second model and a second network interface card (NIC), used to parse the data packets through the second NIC to obtain the data, transmit the data to the second model, and train the second model based on the data.

[0005] Optionally, the first model includes a first physical memory, and the first network interface card includes a first virtual memory; The first device is used to map data in the first physical memory to the first virtual memory through the first model, and to encapsulate the data in the first virtual memory using the first network card based on the video network protocol to obtain the data packet.

[0006] Optionally, the data packet includes the data and a memory write address; the second network card includes a second virtual memory, and the second model includes a second physical memory; The first device is used to send the data packet to the second network card via the first network card based on the video network; The second device is configured to parse the data packet through the second network card to obtain the data and the memory write address, write the data into the second virtual memory according to the memory write address, and map the data in the second virtual memory to the second physical memory.

[0007] Optionally, the second device is configured to determine the current network bandwidth of the network through the second network interface card (NIC), determine whether congestion has occurred based on the network bandwidth, and if congestion has occurred, send a data adjustment request to the first NIC; the data adjustment request includes the network bandwidth. The first device is configured to adjust the transmission rate of the data packet according to the data adjustment request via the first network interface card.

[0008] Optionally, the second device is configured to determine whether the sequence number of the data packet is abnormal through the second network interface card (NIC), and if the sequence number of the data packet is abnormal, send a retransmission request to the first NIC; the retransmission request includes the sequence number of the data packet where the abnormality occurred. The first device is configured to retransmit the data packet according to the retransmission request via the first network interface card.

[0009] This invention also discloses a VRB-based data transmission method applied to a first device, the first device including a first model and a first network interface card (NIC); the first device is communicatively connected to a second device, and the method includes: The data packets are obtained by encapsulating the data used by the first model using the first network interface card based on the video network protocol; The data packet is sent to the second device via the visual network, so that the second device can parse the data packet to obtain the data and perform training based on the data.

[0010] This invention also discloses a VRB-based data transmission method applied to a second device, the second device including a second model and a second network interface card; a first device and the second device are communicatively connected; the method includes: The second network card receives data packets sent by the first device. The data packet is parsed using the second network interface card to obtain the data; The data is transmitted to the second model; The second model is trained based on the data.

[0011] The present invention also discloses a VRB-based data transmission device applied to a first device, the first device including a first model and a first network interface card; the first device is communicatively connected to a second device, the device comprising: The encapsulation module is used to encapsulate the data used by the first model based on the video network protocol through the first network card to obtain data packets; The sending module is used to send the data packet to the second device based on the visual network, so that the second device can parse the data packet to obtain the data and perform training based on the data.

[0012] The present invention also discloses a VRB-based data transmission device applied to a second device, the second device including a second model and a second network interface card; a first device and the second device are communicatively connected; the device includes: A receiving module is used to receive data packets sent by the first device through the second network card; The parsing module is used to parse the data packets through the second network interface card to obtain data; A transmission module is used to transmit the data to the second model; A training module is used to train the second model based on the data.

[0013] The present invention also discloses an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the VRB-based data transmission method described above.

[0014] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the VRB-based data transmission method described above.

[0015] The embodiments of the present invention have the following advantages: This invention discloses a VRB-based data transmission system, method, apparatus, device, and medium. It establishes a visual network communication connection between a first device and a second device. The first network interface card (NIC) of the first device encapsulates the data used by the first model based on the visual network protocol and generates data packets. These data packets are then sent to the second device via the visual network. Finally, the second network interface card of the second device parses the data packets to extract the data and transmits it to the second model for training. This entire process bypasses the significant CPU resource consumption of traditional network protocol stacks, effectively avoiding the problem of computational resource hogging, reducing data transmission latency between models on different devices, and significantly improving data transmission efficiency. It better meets the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, providing stable and efficient transmission support for the second model to efficiently utilize cross-device data for training. Attached Figure Description

[0016] Figure 1This is a structural block diagram of a VRB-based data transmission system provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the steps of a VRB-based data transmission method provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of another VRB-based data transmission method provided in an embodiment of the present invention; Figure 4 This is a structural block diagram of a VRB-based data transmission device provided in an embodiment of the present invention; Figure 5 This is a structural block diagram of another VRB-based data transmission device provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] One of the core concepts of this invention is that it establishes a visual network communication connection between a first device and a second device. The first network card of the first device encapsulates the data used by the first model and generates data packets based on the visual network protocol. These data packets are then sent to the second device via the visual network. Finally, the second network card of the second device parses the data packets to extract the data and transmits it to the second model for training. This entire process bypasses the large consumption of CPU resources by traditional network protocol stacks, effectively avoiding the problem of computing resources being squeezed out, reducing the data transmission latency between models on different devices, and significantly improving data transmission efficiency. This better meets the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, providing stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0019] like Figure 1 The diagram shows a structural block diagram of a VRB-based data transmission system provided by an embodiment of the present invention. The system includes: a first device 101 and a second device 102, which are connected via video network communication. The first device 101 includes a first model 1011 and a first network interface card 1012, which is used to encapsulate the data used by the first model based on the video network protocol to obtain data packets, and send the data packets to the second device based on the video network.

[0020] In this embodiment of the invention, the first device 101 may include two key components: a first model 1011 and a first network card 1012. Its core task is to complete the preprocessing and encapsulation of data and send the data out based on the video network.

[0021] In this stage, the first model 1011 acts as the data source, generating relevant data for its own operation or for subsequent training of the second device 102. This data does not initially meet the format requirements for transmission in the video network, so it needs to be further processed by the first network card 1012.

[0022] The first network interface card 1012 can encapsulate the raw data used by the first model 1011 in accordance with the specifications of the video network protocol. During the encapsulation process, necessary header information, address information, and check information are added to the raw data according to the data packet structure defined by the video network protocol, thereby converting the raw data into data packets that conform to the video network transmission standard. After the encapsulation is completed, the first device 101 can accurately send the encapsulated data packets to the second device 102 with the help of the established video network communication link, ensuring that the data can be transmitted stably in the video network environment.

[0023] The second device 102 includes a second model 1021 and a second network interface card 1022, which is used to parse data packets through the second network interface card, obtain data, transmit the data to the second model, and train the second model based on the data.

[0024] In this embodiment of the invention, the second device 102 may be composed of a second model 1021 and a second network card 1022, and is mainly responsible for receiving data packets sent by the first device, completing data parsing, and using the parsed data for model training.

[0025] When the second device 102 receives a data packet from the first device 101 via the video network, the second network interface card 1022 first parses the data packet. The second network interface card 1022 will reverse-disassemble the data packet according to the corresponding rules of the video network protocol, remove the additional information such as headers, addresses, and checksums added during the encapsulation by the first device, and extract the original data initially generated by the first model 1011. This ensures that the parsed data is consistent with the original data sent by the first device, avoiding data distortion or errors during transmission and parsing. After successfully parsing the original data, the second device 102 will transmit this data to the second model 1021. At this time, the second model 1021 will use the received data as the basis for training, start the model training process, and use this data to adjust its own parameters and optimize the model structure, thereby improving the model performance.

[0026] This invention discloses a VRB-based data transmission system. It establishes a visual network communication connection between a first device and a second device. The first network interface card (NIC) of the first device encapsulates the data used by the first model based on the visual network protocol and generates data packets. These data packets are then sent to the second device via the visual network. Finally, the second network interface card of the second device parses the data packets to extract the data and transmits it to the second model for training. This entire process bypasses the significant CPU resource consumption of traditional network protocol stacks, effectively avoiding the problem of computational resource hogging, reducing data transmission latency between models on different devices, and significantly improving data transmission efficiency. It better meets the stringent requirements of new applications in the rapid iteration of AI technology for data transmission systems, providing stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0027] In one embodiment of the present invention, the first model includes a first physical memory, the first network interface card (NIC) includes a first virtual memory, and the first device is used to map data in the first physical memory to the first virtual memory through the first model, and to encapsulate data in the first virtual memory using the first NIC based on the video network protocol to obtain data packets.

[0028] In embodiments of the present invention, such as Figure 1 The first model 1011 may also include a first physical memory 10111, and the first network interface card 1012 may include a first virtual memory 10121. In the workflow of the first device, the first model uses its own first physical memory to store the data required for operation. This data is the basis for subsequent transmission to the second device. In order to achieve effective data transmission, the first model will map the data in the first physical memory to the first virtual memory contained in the first network interface card. This mapping operation establishes a direct data association between physical memory and virtual memory, enabling data to be efficiently transferred from the physical storage area to the virtual storage area suitable for network processing. Subsequently, the first network interface card 1012 encapsulates the data stored in the first virtual memory 10121 based on the video network protocol, adds various necessary information according to the format structure specified by the protocol, and finally forms a data packet that can be transmitted in the video network, preparing it for subsequent transmission to the second device.

[0029] In one example, in a remote diagnostic application scenario of smart healthcare, the first device can serve as a medical monitoring device deployed at the patient's end. Its first model is an algorithm module responsible for real-time acquisition and preliminary processing of the patient's vital signs data. This continuously generated monitoring data is first stored in the first physical memory of the first model to ensure that the data can be quickly accessed and processed by the model. Since these critical medical data need to be transmitted to a remote second device, the first model can directly map the pre-processed vital signs data in the first physical memory to the first virtual memory of the first network card. This mapping operation avoids duplicate data copying, reduces processing latency, and ensures that the medical data can be ready for transmission at near real-time speed. Subsequently, the first network card encapsulates the vital signs data in the first virtual memory based on the video network protocol, adding necessary content such as patient identification, data acquisition time, and verification information according to the protocol specifications to form a data packet that conforms to the video network transmission standard. This prepares the data for transmission of these critical medical data to the hospital diagnostic server through the video network link, ensuring the real-time performance and reliability of the data during transmission and providing timely and accurate data support for remote diagnosis.

[0030] This invention achieves efficient connection between data from model storage to network transmission preparation through a mapping mechanism between physical memory and virtual memory. This ensures both the speed and accuracy of data transfer, and provides an appropriate operating space for network card protocol encapsulation through virtual memory. In addition, combined with the standardized encapsulation of the video network protocol, it ensures the compatibility and reliability of data transmission in the video network environment, thereby improving the overall processing efficiency and transmission applicability of data from the first model to data packet formation.

[0031] In one embodiment of the present invention, the data packet includes data and a memory write address; the second network interface card (NIC) includes a second virtual memory, and the second model includes a second physical memory; a first device is used to send the data packet to the second NIC via the first NIC based on a video network; the second device is used to parse the data packet via the second NIC to obtain the data and the memory write address, write the data into the second virtual memory according to the memory write address, and map the data in the second virtual memory to the second physical memory.

[0032] In embodiments of the present invention, such as Figure 1The second network interface card 1022 includes a second virtual memory 10221, and the second model 1021 includes a second physical memory 10211. During data transmission, the first network interface card of the first device sends the encapsulated data packet to the second network interface card of the second device through the video network. The data packet not only contains the original data to be transmitted, but also includes memory write address information. After receiving the data packet, the second network interface card parses it and extracts the data and the corresponding memory write address. Then, according to the memory write address, the data is accurately written into the second virtual memory contained in the second network interface card itself, completing the temporary storage of data in the virtual memory. After that, the data stored in the second virtual memory is mapped to the second physical memory contained in the second model, so that the data is finally obtained by the second model and used for training.

[0033] For example, suppose the first model processes a set of image training data. This data is stored in the first physical memory, mapped to the first virtual memory and encapsulated into a data packet. The memory write address corresponding to the image data in the data packet points to a specific area of ​​the second virtual memory. After the second network card parses the data, it writes the image data to that area. Then, through mapping, the second physical memory obtains this image data for the second model to carry out image recognition-related training.

[0034] This invention achieves precise positioning and efficient flow of data from transmission to final storage by carrying memory write addresses in data packets and combining a mapping mechanism between virtual memory and physical memory. This ensures the accuracy of data write location and avoids storage chaos. It also optimizes the connection efficiency of data from network reception to model use by using virtual memory as an intermediate link. At the same time, it ensures the stability of data transmission by relying on video network transmission, thus improving the accuracy and efficiency of cross-device data transmission and application as a whole.

[0035] In one embodiment of the present invention, the second device is used to determine the current network bandwidth of the network through the second network interface card (NIC), determine whether congestion has occurred based on the network bandwidth, and if congestion has occurred, send a data adjustment request to the first NIC; the data adjustment request includes the network bandwidth; the first device is used to adjust the data packet transmission rate through the first NIC according to the data adjustment request.

[0036] In this embodiment of the invention, during the data reception process, the second network interface card (NIC) of the second device continuously monitors the real-time status of the network, continuously monitors and determines the actual bandwidth of the current network. By comparing and analyzing the real-time monitored network bandwidth with a preset normal bandwidth threshold, the second NIC can accurately determine whether network congestion has occurred. When the actual bandwidth is lower than the normal threshold and cannot be recovered for a period of time, it is determined that network congestion has occurred. Once network congestion is confirmed, the second device can send a data adjustment request to the first NIC of the first device through the second NIC. This request contains the actual bandwidth information of the current network so that the first device can make targeted adjustments according to the actual situation. After receiving the data adjustment request through the first NIC, the first device will parse the network bandwidth information and dynamically adjust the data packet sending rate based on it. Usually, the sending rate is reduced to match the actual carrying capacity of the current network, avoiding the aggravation of network congestion due to excessively high sending rates, and ensuring that data transmission can continue under the current network conditions.

[0037] Taking a remote high-definition video conferencing scenario as an example, the first device is the video acquisition device at the conference initiator, responsible for encapsulating real-time captured video data into data packets and sending them to the second device, which is the conference receiver, via the video network. During the conference, the second network card of the second device monitors the network bandwidth in real time. When it detects that the bandwidth has suddenly dropped due to network fluctuations and is lower than the bandwidth threshold required for normal transmission of high-definition video, it determines that network congestion has occurred and then sends a data adjustment request containing the current low bandwidth information to the first network card of the first device. After receiving the request, the first network card of the first device automatically reduces the transmission rate of video data packets according to the bandwidth data, for example, from the original 30 frames per second to 15 frames per second, so that the transmission rate adapts to the current network bandwidth, avoids the accumulation of video data in the network, ensures that the conference video can be transmitted continuously and smoothly, and reduces the phenomenon of lag.

[0038] This invention achieves intelligent perception and adaptive adjustment of network congestion by monitoring network bandwidth in real time and dynamically adjusting the data transmission rate. It can maintain high transmission efficiency when the network conditions are good, and avoid further congestion by reducing the transmission rate when the network is congested. This ensures the stability and continuity of data transmission, improves the adaptability and reliability of the entire data transmission system in complex network environments, and provides a smooth experience guarantee for various application scenarios that rely on stable data transmission.

[0039] In one embodiment of the present invention, the second device is used to determine whether the sequence number of the data packet is abnormal through the second network interface card (NIC). If the sequence number of the data packet is abnormal, a retransmission request is sent to the first NIC. The retransmission request includes the sequence number of the data packet where the abnormality occurred. The first device is used to retransmit the data packet through the first NIC according to the retransmission request.

[0040] In this embodiment of the invention, when the second network card receives a data packet from the first device through the video network, it first extracts the packet sequence number carried in each data packet. These sequence numbers are assigned sequentially by the first device according to the transmission order when encapsulating the data packets, and are used to identify the order in which the data packets are sent, so as to ensure that the second device can reassemble the data in the correct order.

[0041] The second network interface card (NIC) compares the sequence number of the currently received data packet with the sequence number of the previously received data packet to determine if the sequence numbers are consecutive. If the difference between the current sequence number and the previous sequence number is 1, it means that the data packets arrived in order and there is no abnormality. If the difference is greater than 1 or the sequence number is less than the previous sequence number, it is determined that the data packet sequence number is abnormal, which means that there is a data packet loss or a disordered transmission order. Once an abnormal sequence number is detected, the second NIC will immediately generate a retransmission request. This request will clearly indicate the sequence number of the abnormal data packet, and then send the retransmission request to the first NIC of the first device through the video network.

[0042] Upon receiving a retransmission request, the first network interface card (NIC) of the first device first parses the request content and accurately identifies the sequence number of the data packet that needs to be retransmitted. Then, the first NIC searches for the corresponding data packet in its local data packet buffer area based on the abnormal sequence number. This area is used to temporarily store data packets that have been sent but have not been confirmed as successfully received. These data packets are temporarily retained after the initial transmission to handle possible retransmission needs. After finding the target data packet, the first NIC can re-encapsulate these data packets according to the specifications of the video network protocol, ensuring that the encapsulation format is consistent with the initial transmission to avoid new transmission problems. Subsequently, based on the video network, these data packets are retransmitted to the second NIC of the second device. After receiving the retransmitted data packets, the second NIC will verify the sequence number again to confirm that it matches the sequence number in the retransmission request. Then, it will insert these data packets into the data sequence in the correct order to ensure that the data subsequently transmitted to the second model is complete and ordered.

[0043] In a remote industrial control scenario, suppose the first device is a factory's production data acquisition terminal, responsible for collecting real-time equipment operating parameters on the production line, such as machine tool speed, temperature, and pressure. It then sends this data to the second device, which acts as the control center, at a frequency of 10 data packets per second. Each data packet's sequence number starts from 1 and increments sequentially. During a transmission, the second device's second network interface card (NIC) receives a data packet with sequence number 5, but the next received packet is with sequence number 7, leaving a missing data packet with sequence number 6. The second NIC determines this is an anomaly and immediately generates a retransmission request containing sequence number 6, sending it to the first device. The first device's NIC parses the request, finds the data packet with sequence number 6 in its local cache (this packet records the temperature data of a machine tool at a specific moment), repackages it, and sends it to the second device. Upon receiving the retransmitted data packet with sequence number 6, the second NIC inserts it between sequences 5 and 7, ensuring that the complete sequence of equipment operating parameters is transmitted to the second model. Based on this complete data, the second model accurately determines the machine tool's operating status, preventing misjudgments leading to shutdowns or production accidents due to missing data.

[0044] The above process constructs a data transmission integrity guarantee mechanism through real-time detection and anomaly handling of data packet sequence numbers: on the one hand, the sequence number detection of the second network card can quickly detect data packet loss or disorder, preventing abnormal data from entering subsequent processing stages; on the other hand, the accurate retransmission of the first network card based on sequence numbers can selectively supplement missing data without retransmitting all data packets, reducing network bandwidth waste. This mechanism effectively improves the accuracy and integrity of data transmission, ensuring that critical data is not lost or disordered even when the video network experiences occasional fluctuations. It provides reliable transmission guarantees for application scenarios that rely on continuous and complete data, such as industrial control, telemedicine, and high-definition live streaming, while reducing the risk of business interruption due to data anomalies.

[0045] This invention discloses a system for data transmission based on VRB. By establishing a visual network communication connection between a first device and a second device, the first network interface card (NIC) of the first device encapsulates the data used by the first model and generates data packets based on the visual network protocol. These data packets are then sent to the second device via the visual network. Finally, the second network interface card of the second device parses the data packets to extract the data and transmits it to the second model for training. This entire process bypasses the significant CPU resource consumption of traditional network protocol stacks, effectively avoiding the problem of computational resource hogging, reducing data transmission latency between models on different devices, and significantly improving data transmission efficiency. It better meets the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, providing stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0046] Reference Figure 2This diagram illustrates a flowchart of a VRB-based data transmission method according to an embodiment of the present invention, applied to a first device, the first device including a first model and a first network interface card (NIC); the first device is communicatively connected to a second device, the method including: Step 201: The data used by the first model is encapsulated using the first network card based on the video network protocol to obtain data packets; Step 202: Send data packets to the second device based on the visual network, so that the second device can parse the data packets to obtain data and perform training based on the data.

[0047] In one embodiment of the present invention, the first model includes a first physical memory, and the first network interface card (NIC) includes a first virtual memory; the data used by the first model is encapsulated by the first NIC based on the video network protocol to obtain a data packet, including: mapping data in the first physical memory to the first virtual memory through the first model; and encapsulating data in the first virtual memory by the first NIC based on the video network protocol to obtain a data packet.

[0048] In one embodiment of the present invention, the method further includes: receiving a data adjustment request sent by a second network card through a first network card; the data adjustment request includes network bandwidth; and adjusting the data packet transmission rate through the first network card according to the data adjustment request.

[0049] In one embodiment of the present invention, the method further includes: receiving a data retransmission request sent by a second network card through a first network card, the data retransmission request including the sequence number of the data packet where the data packet is abnormal; and retransmitting the data packet according to the retransmission request.

[0050] This invention discloses a VRB-based data transmission method. By establishing a visual network communication connection between a first device and a second device, the first network interface card (NIC) of the first device encapsulates the data used by the first model based on the visual network protocol and generates data packets. These data packets are then sent to the second device via the visual network. Finally, the second network interface card of the second device parses the data packets to extract the data and transmits it to the second model for training. This entire process bypasses the significant CPU resource consumption of traditional network protocol stacks, effectively avoiding the problem of computational resource hogging, reducing data transmission latency between models on different devices, and significantly improving data transmission efficiency. It better meets the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, providing stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0051] Reference Figure 3 This diagram illustrates a flowchart of a VRB-based data transmission method provided by an embodiment of the present invention, applied to a second device, the second device including a second model and a second network interface card; a first device and the second device are communicatively connected; the method includes: Step 301: Receive data packets sent by the first device through the second network card; Step 302: Parse the data packets using the second network card to obtain the data; Step 303: Transfer the data to the second model; Step 304: Train the second model based on the data.

[0052] In one embodiment of the present invention, the data packet includes data and a memory write address; the second network card includes a second virtual memory, and the second model includes a second physical memory; transmitting data to the second model includes: writing data into the second virtual memory according to the memory write address obtained by the second network card, and mapping the data in the second virtual memory to the second physical memory.

[0053] In one embodiment of the present invention, the method further includes: determining the current network bandwidth of the network through the second network card, determining whether congestion has occurred based on the network bandwidth, and if congestion has occurred, sending a data adjustment request to the first network card so that the first network card adjusts the data packet transmission rate according to the data adjustment request; the data adjustment request includes the network bandwidth.

[0054] In one embodiment of the present invention, the method further includes: determining whether the sequence number of the data packet is abnormal through the second network card; if the sequence number of the data packet is abnormal, sending a retransmission request to the first network card so that the first network card retransmits the data packet according to the retransmission request; the retransmission request includes the sequence number of the data packet in which the abnormality occurred.

[0055] This invention discloses a VRB-based data transmission method. By establishing a visual network communication connection between a first device and a second device, the first network interface card (NIC) of the first device encapsulates the data used by the first model based on the visual network protocol and generates data packets. These data packets are then sent to the second device via the visual network. Finally, the second network interface card of the second device parses the data packets to extract the data and transmits it to the second model for training. This entire process bypasses the significant CPU resource consumption of traditional network protocol stacks, effectively avoiding the problem of computational resource hogging, reducing data transmission latency between models on different devices, and significantly improving data transmission efficiency. It better meets the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, providing stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0056] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0057] Reference Figure 4 This diagram illustrates a structural block diagram of a VRB-based data transmission device according to an embodiment of the present invention. The device is applied to a first device, which includes a first model and a first network interface card (NIC). The first device is communicatively connected to a second device. The device includes: The encapsulation module 401 is used to encapsulate the data used by the first model through the first network card based on the video network protocol to obtain data packets; The sending module 402 is used to send data packets to the second device based on the visual network, so that the second device can parse the data packets to obtain data and perform training based on the data.

[0058] This invention discloses a VRB-based data transmission device. By establishing a visual network communication connection between a first device and a second device, the first network card of the first device encapsulates the data used by the first model and generates data packets based on the visual network protocol. These data packets are then sent to the second device via the visual network. Finally, the second network card of the second device parses the data packets to extract the data and transmits it to the second model for training. The entire process bypasses the large consumption of CPU resources by traditional network protocol stacks, effectively avoiding the problem of computing resources being squeezed out, reducing the data transmission latency between models on different devices, and significantly improving data transmission efficiency. It can better meet the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, and provides stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0059] Reference Figure 5 This diagram illustrates a structural block diagram of another VRB-based data transmission device provided in an embodiment of the present invention, applied to a second device. The second device includes a second model and a second network interface card (NIC). A first device and the second device are communicatively connected. The device includes: The receiving module 501 is used to receive data packets sent by the first device through the second network card; The parsing module 502 is used to parse data packets through the second network card to obtain data; Transmission module 503 is used to transmit data to the second model; Training module 504 is used to train the second model based on the data.

[0060] This invention discloses a VRB-based data transmission device. By establishing a visual network communication connection between a first device and a second device, the first network card of the first device encapsulates the data used by the first model and generates data packets based on the visual network protocol. These data packets are then sent to the second device via the visual network. Finally, the second network card of the second device parses the data packets to extract the data and transmits it to the second model for training. The entire process bypasses the large consumption of CPU resources by traditional network protocol stacks, effectively avoiding the problem of computing resources being squeezed out, reducing the data transmission latency between models on different devices, and significantly improving data transmission efficiency. It can better meet the stringent requirements of new applications under the rapid iteration of AI technology for data transmission systems, and provides stable and efficient transmission support for the second model to efficiently utilize cross-device data for training.

[0061] This invention also provides an electronic device, comprising: It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the various processes of the above-described VRB-based data transmission method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0062] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described VRB-based data transmission method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0063] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0064] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0066] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0068] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0069] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0070] The present invention has provided a detailed description of a VRB-based data transmission system, method, apparatus, device, and storage medium. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A VRB-based data transmission system, characterized in that, The system includes: a first device and a second device, wherein the first device and the second device are connected via video network communication. The first device includes a first model and a first network interface card (NIC), and is used to encapsulate the data used by the first model based on the video network protocol to obtain a data packet, and send the data packet to the second device based on the video network protocol; The second device includes a second model and a second network interface card (NIC), used to parse the data packet through the second NIC to obtain the data, transmit the data to the second model, and train the second model based on the data.

2. The VRB-based data transmission system according to claim 1, characterized in that, The first model includes a first physical memory, and the first network card includes a first virtual memory; The first device is used to map data in the first physical memory to the first virtual memory through the first model, and to encapsulate the data in the first virtual memory through the first network card based on the video network protocol to obtain the data packet.

3. The VRB-based data transmission system according to claim 1, characterized in that, The data packet includes the data and a memory write address; the second network card includes a second virtual memory, and the second model includes a second physical memory; The first device is used to send the data packet to the second network card via the first network card based on the video network; The second device is configured to parse the data packet through the second network card to obtain the data and the memory write address, write the data into the second virtual memory according to the memory write address, and map the data in the second virtual memory to the second physical memory.

4. The VRB-based data transmission system according to claim 1, characterized in that, The second device is used to determine the current network bandwidth through the second network interface card (NIC), determine whether congestion has occurred based on the network bandwidth, and if congestion has occurred, send a data adjustment request to the first NIC; the data adjustment request includes the network bandwidth. The first device is configured to adjust the transmission rate of the data packet according to the data adjustment request via the first network interface card.

5. A data transmission method based on VRB, characterized in that, The method is applied to a first device, the first device including a first model and a first network interface card (NIC); the first device is communicatively connected to a second device, the method comprising: The data packets are obtained by encapsulating the data used by the first model using the first network interface card based on the video network protocol; The data packet is sent to the second device via the visual network, so that the second device can parse the data packet to obtain the data and perform training based on the data.

6. A data transmission method based on VRB, characterized in that, The method is applied to a second device, which includes a second model and a second network interface card (NIC); the first device and the second device are communicatively connected; the method includes: The second network card receives data packets sent by the first device. The data packet is parsed using the second network interface card to obtain the data; The data is transmitted to the second model; The second model is trained based on the data.

7. A data transmission device based on VRB, characterized in that, Applied to a first device, the first device including a first model and a first network interface card; the first device is communicatively connected to a second device, the device comprising: The encapsulation module is used to encapsulate the data used by the first model based on the video network protocol through the first network card to obtain data packets; The sending module is used to send the data packet to the second device based on the visual network, so that the second device can parse the data packet to obtain the data and perform training based on the data.

8. A data transmission device based on VRB, characterized in that, Applied to a second device, the second device including a second model and a second network interface card; the first device and the second device are communicatively connected; the device includes: A receiving module is used to receive data packets sent by the first device through the second network card; The parsing module is used to parse the data packets through the second network interface card to obtain data; A transmission module is used to transmit the data to the second model; A training module is used to train the second model based on the data.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the VRB-based data transmission method as described in any one of claims 5 or 6.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the VRB-based data transmission method as described in any one of claims 5 or 6.

Images