CN116527510A - Model transmission method, device, electronic equipment and readable storage medium - Google Patents
Model transmission method, device, electronic equipment and readable storage medium Download PDFInfo
- Publication number
- CN116527510A CN116527510A CN202210065202.9A CN202210065202A CN116527510A CN 116527510 A CN116527510 A CN 116527510A CN 202210065202 A CN202210065202 A CN 202210065202A CN 116527510 A CN116527510 A CN 116527510A
- Authority
- CN
- China
- Prior art keywords
- model
- transmission
- slices
- slice
- path
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005540 biological transmission Effects 0.000 title claims abstract description 283
- 238000000034 method Methods 0.000 title claims abstract description 55
- 230000006870 function Effects 0.000 claims description 23
- 230000011218 segmentation Effects 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 12
- 239000004973 liquid crystal related substance Substances 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 4
- 238000012546 transfer Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 24
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000012545 processing Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 4
- 239000002699 waste material Substances 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The present invention relates to the field of communications technologies, and in particular, to a model transmission method, a device, an electronic apparatus, and a readable storage medium. The specific implementation scheme is as follows: acquiring n transmission paths between a transmitting end node and a receiving end node, wherein n is a positive integer greater than 1; dividing a model to be transmitted in a transmitting end node into n groups of model slices according to the transmission capacity of each transmission path in proportion, and forming a path selection strategy which corresponds the n groups of model slices to the n transmission paths one by one; and transmitting the segmented n groups of model slices to a receiving end node through corresponding transmission paths respectively. According to the invention, the model is segmented and then transmitted through different transmission paths, so that the transmission time delay can be remarkably reduced, and the model transmission efficiency can be improved.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a model transmission method, a device, an electronic apparatus, and a readable storage medium.
Background
In future everything intelligent network, network nodes tend to be intelligent, network node intellectualization causes rapid expansion of information space and even dimension disaster, aggravates difficulty in representing information bearing space, causes difficulty in matching traditional network service capacity and high-dimension information space, and causes overlarge data volume of communication transmission, so that an information service system cannot continuously meet the requirements of complex, diversified and intelligent information transmission of people. And the service information is encoded, transmitted and decoded through the artificial intelligent model, so that the data transmission quantity in the communication service can be obviously reduced, and the information transmission efficiency is greatly improved. These models are relatively stable and have reusability and transmissibility. The propagation and multiplexing of the model are beneficial to enhancing the network intelligence, reducing the cost and the resource waste, and forming an intelligent network with intelligent nodes and extremely simple network.
The core of the intelligent network is a transmission model, the network has a storage function, and the model is stored in the network, possibly at the end user side or at the cloud. Each node can absorb the models of other nodes on the network to realize self-evolution, so that the efficiency of the propagation model directly determines the communication efficiency. However, since some models are larger, the delay of the model transmission is larger and the communication efficiency is reduced by transmitting the models through the same path, an efficient model transmission method and device are needed.
Disclosure of Invention
The invention provides a method, a device, equipment and a storage medium for transmitting a model in an intelligent network.
According to an aspect of the present invention, there is provided a model transmission method including:
acquiring n transmission paths between a transmitting end node and a receiving end node, wherein n is a positive integer greater than 1;
dividing a model to be transmitted in the sending end node into n groups of model slices according to the transmission capacity of each transmission path in proportion, and forming a path selection strategy which corresponds the n groups of model slices to the n transmission paths one by one;
and transmitting the divided n groups of model slices to the receiving end node through the corresponding transmission paths.
Optionally, the dividing the model to be transmitted in the sending end node into n groups of model slices according to the proportion includes:
the method comprises the steps of obtaining total transmission delay as an objective function by modeling the splitting delay of splitting the model to be transmitted into the model slices and the transmission delay of each model slice on the corresponding transmission path;
and obtaining a model segmentation proportion and the path selection strategy by solving and minimizing the objective function, and segmenting the model to be transmitted into n groups of model slices according to the model segmentation proportion.
Optionally, each set of said model slices comprises one or more of said model slices.
Optionally, the model transmission method further includes:
before transmitting each model slice, determining whether the model slice is already stored in any routing node of the current transmission path:
if yes, switching to the next transmission path for transmission, and storing the current model slice in the corresponding routing node in the transmission process;
if not, the current transmission path is used for transmission, and the current model slice is stored in the corresponding routing node in the transmission process.
Optionally, the transmission capability includes a channel capacity of the transmission path.
According to another aspect of the present invention, there is provided a model transmission apparatus including:
the path acquisition module is used for acquiring n transmission paths between the sending end node and the receiving end node, wherein n is a positive integer greater than 1;
the model segmentation module is used for segmenting a model to be transmitted in the sending end node into n groups of model slices according to the transmission capacity of each transmission path in proportion, and forming a path selection strategy for enabling the n groups of model slices to correspond to the n transmission paths one by one;
and the model slice transmission module is used for respectively transmitting the divided n groups of model slices to the receiving end node through the corresponding transmission paths.
Optionally, the process of the model slicing module to slice the model to be transmitted in the sender node into n groups of model slices according to the proportion includes:
the method comprises the steps of obtaining total transmission delay as an objective function by modeling the splitting delay of splitting the model to be transmitted into the model slices and the transmission delay of each model slice on the corresponding transmission path;
and obtaining a model segmentation proportion and the path selection strategy by solving and minimizing the objective function, and segmenting the model to be transmitted into n groups of model slices according to the model segmentation proportion.
Optionally, each set of the model slices cut by the model slicing module includes one or more of the model slices.
Optionally, the method further comprises:
a path selection module, configured to determine, before transmitting each of the model slices, whether the model slice is already stored in any routing node of the current transmission path:
responding to the model slice already stored in the corresponding transmission path, switching to the next transmission path by the model slice transmission module for transmission, and storing the current model slice in the corresponding routing node in the transmission process;
and in response to the model slice not being stored in the corresponding transmission path, the model slice transmission module uses the current transmission path to transmit and stores the current model slice in the corresponding routing node in the transmission process.
Optionally, the transmission capability includes a channel capacity of the transmission path.
The invention also provides an electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the model transmission method of any one of the above-described aspects.
The invention also provides a non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the model transmission method according to any one of the above embodiments.
The invention also provides a computer program product comprising a computer program which, when executed by a processor, implements a model transmission method according to any of the above embodiments.
According to the model transmission method, device, electronic equipment and storage medium in the technical scheme, the model is divided and then transmitted through different transmission paths, so that the transmission time delay can be remarkably reduced, and the model transmission efficiency is improved.
It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the invention or to delineate the scope of the invention. Other features of the present invention will become apparent from the description that follows.
Drawings
The drawings are included to provide a better understanding of the present invention and are not to be construed as limiting the invention. Wherein:
FIG. 1 is a flow chart of the steps of a model transmission method in an embodiment of the invention;
FIG. 2 is a schematic representation of a first model slice transmission in an embodiment of the invention;
FIG. 3 is a schematic representation of a second model slice transmission in an embodiment of the invention;
FIG. 4 is a second exemplary model slice transmission diagram in accordance with an embodiment of the present invention;
FIG. 5 is a second exemplary model slice transmission diagram in accordance with an embodiment of the present invention;
FIG. 6 is a functional block diagram of a first model transmission device in an embodiment of the invention;
FIG. 7 is a functional block diagram of a second model transmission device in an embodiment of the invention;
fig. 8 is a model slice propagation flowchart of a model transmission method in an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding, and are to be considered merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
The invention discloses a method for transmitting service information in an intelligent network mainly through an artificial intelligent model, wherein the first service information to be transmitted is compressed into second service information related to the artificial intelligent model through the artificial intelligent model, so that the data traffic in the network is greatly reduced, and the compression efficiency is far higher than that of a traditional compression algorithm. The sending end equipment extracts the first service information by utilizing a preconfigured first model and obtains second service information to be transmitted; and the sending end equipment transmits the second service information to the receiving end equipment. The receiving terminal equipment receives the second service information and carries out recovery processing on the second service information by utilizing a second pre-configured model to obtain third service information; the third service information recovered by the second model has a slight quality difference compared with the original first service information, but the third service information and the first service information are consistent in content, and the experience of the user is almost unchanged. Before the sending end device transmits the second service information to the receiving end device, the method further comprises: the updating module judges whether the receiving end equipment needs to update the second model, and transmits a preconfigured third model to the receiving end equipment when judging that the second model needs to be updated, and the receiving end equipment updates the second model by using the third model. The service information is processed through the pre-trained artificial intelligent model, so that the data transmission quantity in the communication service can be obviously reduced, and the information transmission efficiency is greatly improved. These models are relatively stable and have reusability and transmissibility. The propagation and multiplexing of the model will help to enhance network intelligence while reducing overhead and resource waste. The model can be divided into a plurality of model slices according to different dividing rules, the model slices can be transmitted among different network nodes, and the model slices can be assembled into the model. Model slices may be stored scattered across multiple network nodes. When a network node requests to find itself missing or needing to update a model or a slice of a model, it may request from surrounding nodes that may have the slice by way of a request.
And transmitting the service information and the model occur in a network layer of a communication network, and communication transmission is performed based on a network layer protocol. The network nodes passing on the path transmitting the traffic information and the model comprise intelligent Jian Lu routers. The functions of the intelligent Jian Lu router include, but are not limited to, business information transmission, model transmission, absorption model self-updating, security protection and the like. The transmission function of the intelligent Jian Lu router involves transmitting traffic information or models from a source node to a sink node, where there are multiple paths between the source node and the sink node. The model transmission function of the intelligent Jian Lu router can transmit model slices, and the model slices are multiplexed by reasonably arranging the model slices to travel a plurality of paths, so that the model transmission rate is improved.
The invention also discloses a model transmission method, as shown in fig. 1, comprising the following steps:
step S101, n transmission paths between a transmitting end node and a receiving end node are obtained, wherein n is a positive integer greater than 1;
step S102, dividing a model to be transmitted in a transmitting end node into n groups of model slices according to the transmission capacity of each transmission path in proportion, and forming a path selection strategy for enabling the n groups of model slices to correspond to the n transmission paths one by one;
step S103, the divided n groups of model slices are transmitted to the receiving end nodes through corresponding transmission paths respectively.
Specifically, the core of the model transmission method is that the model to be transmitted is divided into a plurality of groups of model slices and then is transmitted simultaneously through different transmission paths, and compared with the mode that a single transmission path transmits the whole model, the mode that a plurality of transmission paths transmit the model slices simultaneously is reasonably arranged, so that the transmission delay can be effectively reduced, and the communication efficiency is improved.
Illustratively, assuming node a is a transmitting node and node B is a receiving node, there are n transmission paths with model slice transmission capability between nodes A, B. When the node A needs to send the model to the node B, the model is not transmitted from one path, but the deep learning model is segmented into n groups according to the data quantity as shown in fig. 8, the n groups of slices are transmitted on a plurality of paths, and a plurality of model slices are ensured to arrive at a sink at the same time. For example, as shown in fig. 2, 2 transmission paths, namely transmission paths L1 and L2, are provided between the nodes A, B, and the model to be transmitted is divided into 2 model slices, wherein the slices 1 and 2 are transmitted through the paths L1 and L2 respectively; meanwhile, the sizes of the slices 1 and 2 may be divided according to the transmission capability of the transmission paths L1, L2, which may include the channel capacity of the transmission paths, i.e., the maximum transmission rate of the channels. For example, if the channel capacities of the paths L1 and L2 are the same or have little difference, the model can be transmitted after being divided equally, and the transmission delay can be reduced to half of the original delay; assuming that the channel capacity of the transmission path L1 is 2 times that of the transmission path L2, the model may be sliced according to the ratio of 2:1, the size of slice 1 after slicing is 1/3 of the model, the size of slice 2 is 2/3 of the model, slice 1 is transmitted through the transmission path L2 with smaller channel capacity, slice 2 is transmitted through the transmission path L1 with larger channel capacity, and the efficiency of model transmission can be maximized by slicing the model and reasonably distributing the transmission paths of the model slices.
It should be noted that the above embodiment is only an alternative implementation, and the core of the present invention is to reduce the transmission delay by splitting and separately transmitting the model, and the splitting method of the model is not limited to the above ratio. Each set of model slices may be one model slice or a plurality of model slices. For example, the model may be sliced into 10 model slices, and slices 1-5 may be transmitted through one transmission path and slices 6-10 through another transmission path assuming 2 transmission paths are present.
As an alternative embodiment, the splitting the model to be transmitted in the sender node into n groups of model slices according to the proportion includes:
splitting delay t for splitting a model to be transmitted into model slices 1 And the transmission time delay t of each model slice on the corresponding transmission path 2 Modeling is carried out to obtain the total transmission delay as an objective function;
and obtaining a model segmentation proportion and a path selection strategy by solving the minimized objective function, and segmenting the model to be transmitted into n groups of model slices according to the model segmentation proportion.
Specifically, in this embodiment, the objective function min { t } can be obtained by modeling 1 +t 2 It is assumed that there is a matrix N in which the data is the path number (1, 2, …, N) taken by each slice (1, 2, …, N) in one-to-one correspondence, K being the slicing ratio, e.g. [0.12,0.32,0.24,0.32 ]]According to the modeling model, the segmentation time delay t can be calculated according to K and N 1 And transmission delay t 2 . Specifically, through deep learning modeling, the input parameters of the deep neural network are set to be the segmentation proportion K and the matrix N, and the output parameters are set to be t 1 +t 2 Finding out the optimal K and N through multiple training so as to enable the objective function min { t } 1 +t 2 The minimum value of the method, namely the minimum total transmission delay of the model, maximizes the transmission capacity of each transmission path, and improves the transmission efficiency of the model by using limited communication resources.
As an alternative embodiment, the model transmission method further includes:
before transmitting each model slice, it is determined whether the model slice is already stored in any one of the routing nodes of the current transmission path:
if yes, switching to the next transmission path for transmission, and storing the current model slice in the corresponding routing node in the transmission process, wherein the step is that the partial model slice changes the transmission path and is not completely consistent with the preliminarily determined path selection strategy, so that the transmission capacity of the transmission path is possibly not matched, and each routing node is required to be synchronous, and the model slice can reach the receiving end node simultaneously;
if not, the current transmission path is used for transmission, and the current model slice is stored in the corresponding routing node in the transmission process.
Specifically, after the model to be transmitted is sliced, it is possible that some routing nodes of the transmission path have stored some model slices of the model, while other routing nodes of the transmission path have not stored the model slices. For example, as shown in fig. 3, assuming that the model to be transmitted is sliced into 10 model slices, there are slices 1-10 in the sender node a, while the receiver node B has only slices 1-7, thus requiring the sender node a to transmit its required slices 8, 9, 10. At this time, the transmission paths L1 and L2 between the nodes a and B can transmit, and it can be seen in fig. 3 that the routing node C in the first transmission path L1 has backed up the slices 1-5, the routing node D has backed up the slices 1-5, the slice 8, and no backup slices 9 and 10; the routing node E of the second transmission path L2 backs up the slices 1-4, the routing node F backs up the slices 1-3 and 6-9, and no backup slice 8 is provided; the slices 9, 10 required by the receiving end node B can then be transmitted via the first transmission path L1 and the slice 8 via the second transmission path L2. In the embodiment, 3 model slices are divided into 2 groups of model slices, and the model slices are transmitted simultaneously through 2 transmission paths, so that the transmission delay can be effectively reduced; in addition, when the model slice is transmitted, a transmission path without storing the model slice is selected as far as possible, so that the model slice can be shared with a routing node without backing up the model slice, and the useful model slice is stored in the routing node, thereby improving the sharing property of the network.
Further, as shown in fig. 4, when the slices 9 and 10 are transmitted by using the first transmission path L1, the slices 9 and 10 can be backed up on the routing node (C, D) of the first transmission path L1 at the same time, for example, the node C has stored the slices 1-5 and the node D has stored the slices 1-5 and 8, so that the slices 9 and 10 can be backed up on the nodes C and D, and thus, the node C or D does not need to transmit through the node a again when the slices 9 and 10 are needed, which is beneficial to saving communication resources. Likewise, when transmitting the slice 8 using the second transmission path L2, the slice 8 may be backed up on the node E without the slice 8, and the node F where the slice 8 is backed up does not need to be backed up again. Fig. 5 shows that after model slices 8, 9, 10 reach node B, node B now has all slices 1-10 of the model. The transmission method not only can reduce the transmission delay of the model, but also can improve the sharing property of the model.
Alternatively, slice 8 may be directly transmitted to node B by route node D that has backed up slice 8, slice 9 may be directly transmitted to node B by route node F that has backed up slice 9, and only slice 10 may be transmitted to node B by node a, thereby further improving the transmission efficiency of the model slice and reducing the bandwidth occupied by transmitting the model slice. However, this transmission mode requires that the model has formed a certain sharing property in the network, and by this transmission mode, the sharing property of using the model slice can be maximized.
Illustratively, any one of the model transmission methods of the above embodiments may be applied to the smart medical field. With advances in technology, the medical industry will incorporate more artificial intelligence technology. Along with the construction of intelligent medical treatment, medical services will be truly intelligent. By using extensive medical data, an effective deep learning model is trained, and according to the data of a patient, the condition of the patient can be effectively analyzed and assisted to be diagnosed, so that the patient can enjoy high-quality medical service with shorter treatment time and basic treatment cost. Under the intelligent medical scene, the intelligent network can utilize the advantages of the self-propagation model, when one service node has trained the model, the node does not need to waste a great deal of time to train the model when the other service node has service requirements, the node which has trained the similar model can be directly requested from the network, and the model can be directly transmitted in a model slicing mode, so that a useful slice is reserved in the network node, the model transmission rate is improved, and precious diagnosis and treatment time is saved.
The invention also provides a model transmission device, as shown in fig. 6, comprising:
a path acquisition module 601, configured to acquire n transmission paths between a transmitting end node and a receiving end node, where n is a positive integer greater than 1;
the model slicing module 602 is configured to slice a model to be transmitted in a sender node into n groups of model slices according to a proportion according to transmission capabilities of each transmission path, and form a path selection policy that corresponds the n groups of model slices to the n transmission paths one by one;
and the model slice transmission module 603 is configured to transmit the sliced n groups of model slices to the receiving end nodes through corresponding transmission paths respectively.
Specifically, the core of the model transmission method is that the model to be transmitted is divided into a plurality of groups of model slices and then is transmitted simultaneously through different transmission paths, and compared with the mode that a single transmission path transmits the whole model, the mode that a plurality of transmission paths transmit the model slices simultaneously is reasonably arranged, so that the transmission delay can be effectively reduced, and the communication efficiency is improved. For example, assuming that the node a is a transmitting end node and the node B is a receiving end node, n transmission paths with model slice transmission capability existing between the nodes A, B are firstly obtained through the path obtaining module 601, when the node a needs to send a model to the node B, the model is not transmitted from one path, but is divided into n groups according to the data quantity through the model dividing module 602, the model slice transmission module 603 distributes the n groups of model slices to transmit on the multiple transmission paths, so that the multiple model slices arrive at the receiving end node at the same time, and the advantage of model dividing transmission is that the transmission rate of the model in the network can be improved, and the efficiency of network communication is improved. For example, as shown in fig. 2, 2 transmission paths, namely transmission paths L1 and L2, are provided between the nodes A, B, and the model to be transmitted is divided into 2 model slices, wherein the slices 1 and 2 are transmitted through the paths L1 and L2 respectively; meanwhile, the sizes of the slices 1 and 2 may be divided according to the transmission capability of the transmission paths L1, L2, which may include the channel capacity of the transmission paths, i.e., the maximum transmission rate of the channels. For example, if the channel capacities of the paths L1 and L2 are the same or have little difference, the model can be transmitted after being divided equally, and the transmission delay can be reduced to half of the original delay; assuming that the channel capacity of the transmission path L1 is 2 times that of the transmission path L2, the model may be sliced according to the ratio of 2:1, the size of slice 1 after slicing is 1/3 of the model, the size of slice 2 is 2/3 of the model, slice 1 is transmitted through the transmission path L2 with smaller channel capacity, slice 2 is transmitted through the transmission path L1 with larger channel capacity, and the efficiency of model transmission can be maximized by slicing the model and reasonably distributing the transmission paths of the model slices.
It should be noted that the above embodiment is only an alternative implementation, and the core of the present invention is to reduce the transmission delay by splitting and separately transmitting the model, and the splitting method of the model is not limited to the above ratio. Each set of model slices may be one model slice or a plurality of model slices. For example, the model may be sliced into 10 model slices, and slices 1-5 may be transmitted through one transmission path and slices 6-10 through another transmission path assuming 2 transmission paths are present.
As an alternative implementation, the process of splitting the model to be transmitted in the sender node into n groups of model slices by the model splitting module 602 includes:
modeling is carried out on the splitting time delay of the model to be transmitted and the transmission time delay of each model slice on the corresponding transmission path to obtain the total transmission time delay as an objective function;
and obtaining a model segmentation proportion and a path selection strategy by solving the minimized objective function, and segmenting the model to be transmitted into n groups of model slices according to the model segmentation proportion.
Specifically, the model segmentation module 602 in this embodiment may obtain the objective function min { t } through modeling 1 +t 2 It is assumed that there is a matrix N in which the data is the path number (1, 2, …, N) taken by each slice (1, 2, …, N) in one-to-one correspondence, K being the slicing ratio, e.g. [0.12,0.32,0.24,0.32 ]]According to the modeling model, the segmentation time delay t can be calculated according to K and N 1 And transmission delay t 2 . Specifically, through deep learning modeling, the input parameters of the deep neural network are set to be the segmentation proportion K and the matrix N, and the output parameters are set to be t 1 +t 2 Finding out the optimal K and N through multiple training so as to enable the objective function min { t } 1 +t 2 The minimum value of the method, namely the minimum total transmission delay of the model, maximizes the transmission capacity of each transmission path, and improves the transmission efficiency of the model by using limited communication resources.
As an alternative embodiment, as shown in fig. 7, the model transmission device further includes:
a path selection module 604, configured to determine, before transmitting each model slice, whether the model slice is already stored in any routing node of the current transmission path:
in response to the model slice being already stored in the corresponding transmission path, the model slice transmission module 603 switches to the next transmission path to transmit, and stores the current model slice in the corresponding routing node in the transmission process, where it is to be noted that, because part of the model slice changes the transmission path and is not completely consistent with the initially determined path selection policy, the transmission capability of the transmission path may be mismatched, and thus synchronization of each routing node is required, so that the model slice can reach the receiving end node at the same time;
in response to the model slice not being stored in the corresponding transmission path, the model slice transmission module 603 uses the current transmission path to transmit and stores the current model slice in the corresponding routing node during transmission.
Specifically, after the model to be transmitted is sliced, it is possible that some routing nodes of the transmission path have stored some model slices of the model, while other routing nodes of the transmission path have not stored the model slices. For example, as shown in fig. 3, assuming that the model to be transmitted is sliced into 10 model slices, there are slices 1-10 in the sender node a, while the receiver node B has only slices 1-7, thus requiring the sender node a to transmit its required slices 8, 9, 10. At this time, the transmission paths L1 and L2 between the nodes a and B can transmit, and it can be seen in fig. 3 that the routing node C in the first transmission path L1 has backed up the slices 1-5, the routing node D has backed up the slices 1-5, the slice 8, and no backup slices 9 and 10; the routing node E of the second transmission path L2 backs up the slices 1-4, the routing node F backs up the slices 1-3 and 6-9, and no backup slice 8 is provided; the slices 9, 10 required by the receiving end node B can then be transmitted via the first transmission path L1 and the slice 8 via the second transmission path L2. In the embodiment, 3 model slices are divided into 2 groups of model slices, and the model slices are transmitted simultaneously through 2 transmission paths, so that the transmission delay can be effectively reduced; in addition, the path selection module 604 selects a transmission path that does not store the model slice as far as possible for transmission, so that the model slice can be shared with a routing node that does not backup the model slice, and the useful model slice is stored in the routing node, thereby improving the sharing performance of the network.
Further, as shown in fig. 4, when the slices 9 and 10 are transmitted by using the first transmission path L1, the slices 9 and 10 can be backed up on the routing node (C, D) of the first transmission path L1 at the same time, for example, the node C has stored the slices 1-5 and the node D has stored the slices 1-5 and 8, so that the slices 9 and 10 can be backed up on the nodes C and D, and thus, the node C or D does not need to transmit through the node a again when the slices 9 and 10 are needed, which is beneficial to saving communication resources. Likewise, when transmitting the slice 8 using the second transmission path L2, the slice 8 may be backed up on the node E without the slice 8, and the node F where the slice 8 is backed up does not need to be backed up again. Fig. 5 shows that after model slices 8, 9, 10 reach node B, node B now has all slices 1-10 of the model. The transmission method not only can reduce the transmission delay of the model, but also can improve the sharing property of the model.
Alternatively, slice 8 may be directly transmitted to node B by route node D that has backed up slice 8, slice 9 may be directly transmitted to node B by route node F that has backed up slice 9, and only slice 10 may be transmitted to node B by node a, thereby further improving the transmission efficiency of the model slice and reducing the bandwidth occupied by transmitting the model slice. However, this transmission mode requires that the model has formed a certain sharing property in the network, and by this transmission mode, the sharing property of using the model slice can be maximized.
Illustratively, any of the model transmission apparatuses of the above embodiments may be applied to the smart medical field. With advances in technology, the medical industry will incorporate more artificial intelligence technology. Along with the construction of intelligent medical treatment, medical services will be truly intelligent. By using extensive medical data, an effective deep learning model is trained, and according to the data of a patient, the condition of the patient can be effectively analyzed and assisted to be diagnosed, so that the patient can enjoy high-quality medical service with shorter treatment time and basic treatment cost. Under the intelligent medical scene, the intelligent network can utilize the advantages of the self-propagation model, when one service node has trained the model, the node does not need to waste a great deal of time to train the model when the other service node has service requirements, the node which has trained the similar model can be directly requested from the network, and the model can be directly transmitted in a model slicing mode, so that a useful slice is reserved in the network node, the model transmission rate is improved, and precious diagnosis and treatment time is saved.
According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium and a computer program product.
In particular, electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.
The apparatus includes a computing unit that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) or a computer program loaded from a storage unit into a Random Access Memory (RAM). In the RAM, various programs and data required for the operation of the device may also be stored. The computing unit, ROM and RAM are connected to each other by a bus. An input/output (I/O) interface is also connected to the bus.
A plurality of components in a device are connected to an I/O interface, comprising: an input unit such as a keyboard, a mouse, etc.; an output unit such as various types of displays, speakers, and the like; a storage unit such as a magnetic disk, an optical disk, or the like; and communication units such as network cards, modems, wireless communication transceivers, and the like. The communication unit allows the device to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.
The computing unit may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing units include, but are not limited to, central Processing Units (CPUs), graphics Processing Units (GPUs), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital Signal Processors (DSPs), and any suitable processors, controllers, microcontrollers, and the like. The calculation unit performs the respective methods and processes described above, such as the model transfer method in the above-described embodiment. For example, in some embodiments, the model transmission method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and/or installed onto the device via the ROM and/or the communication unit. One or more steps of the model transmission method described above may be performed when the computer program is loaded into RAM and executed by the computing unit. Alternatively, in other embodiments, the computing unit may be configured to perform the model transmission method by any other suitable means (e.g. by means of firmware).
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for carrying out the model transmission method of the present invention may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of the present invention, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.
The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server incorporating a blockchain.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps described in the present invention may be performed in parallel, sequentially, or in a different order, so long as the desired results of the technical solution disclosed in the present invention can be achieved, and are not limited herein.
The above embodiments do not limit the scope of the present invention. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of the present invention.
Claims (13)
1. A model transmission method, comprising:
acquiring n transmission paths between a transmitting end node and a receiving end node, wherein n is a positive integer greater than 1;
dividing a model to be transmitted in the sending end node into n groups of model slices according to the transmission capacity of each transmission path in proportion, and forming a path selection strategy which corresponds the n groups of model slices to the n transmission paths one by one;
and transmitting the divided n groups of model slices to the receiving end node through the corresponding transmission paths.
2. The model transmission method according to claim 1, wherein the dividing the model to be transmitted in the transmitting end node into n groups of model slices according to the proportion includes:
the method comprises the steps of obtaining total transmission delay as an objective function by modeling the splitting delay of splitting the model to be transmitted into the model slices and the transmission delay of each model slice on the corresponding transmission path;
and obtaining a model segmentation proportion and the path selection strategy by solving and minimizing the objective function, and segmenting the model to be transmitted into n groups of model slices according to the model segmentation proportion.
3. The model transmission method according to claim 1 or 2, characterized in that each set of the model slices comprises one or more of the model slices.
4. A model transmission method according to claim 3, characterized in that the model transmission method further comprises:
before transmitting each model slice, determining whether the model slice is already stored in any routing node of the current transmission path:
if yes, switching to the next transmission path for transmission, and storing the current model slice in the corresponding routing node in the transmission process;
if not, the current transmission path is used for transmission, and the current model slice is stored in the corresponding routing node in the transmission process.
5. The model transmission method according to any one of claims 1 to 4, characterized in that the transmission capability comprises a channel capacity of the transmission path.
6. A model transfer device, comprising:
the path acquisition module is used for acquiring n transmission paths between the sending end node and the receiving end node, wherein n is a positive integer greater than 1;
the model segmentation module is used for segmenting a model to be transmitted in the sending end node into n groups of model slices according to the transmission capacity of each transmission path in proportion, and forming a path selection strategy for enabling the n groups of model slices to correspond to the n transmission paths one by one;
and the model slice transmission module is used for respectively transmitting the divided n groups of model slices to the receiving end node through the corresponding transmission paths.
7. The model transmission apparatus according to claim 6, wherein the process of the model slicing module slicing the model to be transmitted in the transmitting end node into n groups of model slices according to a ratio includes:
the method comprises the steps of obtaining total transmission delay as an objective function by modeling the splitting delay of splitting the model to be transmitted into the model slices and the transmission delay of each model slice on the corresponding transmission path;
and obtaining a model segmentation proportion and the path selection strategy by solving and minimizing the objective function, and segmenting the model to be transmitted into n groups of model slices according to the model segmentation proportion.
8. The model transfer apparatus of claim 6 or 7, wherein each set of the model slices sliced by the model slicing module comprises one or more of the model slices.
9. The model transmission apparatus according to claim 8, further comprising:
a path selection module, configured to determine, before transmitting each of the model slices, whether the model slice is already stored in any routing node of the current transmission path:
responding to the model slice already stored in the corresponding transmission path, switching to the next transmission path by the model slice transmission module for transmission, and storing the current model slice in the corresponding routing node in the transmission process;
and in response to the model slice not being stored in the corresponding transmission path, the model slice transmission module uses the current transmission path to transmit and stores the current model slice in the corresponding routing node in the transmission process.
10. The model transmission apparatus according to any one of claims 6-9, wherein the transmission capability comprises a channel capacity of the transmission path.
11. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the model transmission method of any one of claims 1-5.
12. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the model transmission method according to any one of claims 1-5.
13. A computer program product comprising a computer program which, when executed by a processor, implements the model transmission method according to any one of claims 1-5.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210065202.9A CN116527510A (en) | 2022-01-20 | 2022-01-20 | Model transmission method, device, electronic equipment and readable storage medium |
| PCT/CN2022/136415 WO2023138233A1 (en) | 2022-01-20 | 2022-12-04 | Model transmission method and apparatus, electronic device and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210065202.9A CN116527510A (en) | 2022-01-20 | 2022-01-20 | Model transmission method, device, electronic equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116527510A true CN116527510A (en) | 2023-08-01 |
Family
ID=87347757
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210065202.9A Pending CN116527510A (en) | 2022-01-20 | 2022-01-20 | Model transmission method, device, electronic equipment and readable storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN116527510A (en) |
| WO (1) | WO2023138233A1 (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105847139A (en) * | 2016-03-25 | 2016-08-10 | 乐视控股（北京）有限公司 | Data transmission method, apparatus and system |
| US10958579B2 (en) * | 2019-05-14 | 2021-03-23 | Vmware, Inc. | Congestion avoidance in a slice-based network |
| CN110022261A (en) * | 2019-05-20 | 2019-07-16 | 北京邮电大学 | Multi-path transmission method and apparatus based on SCTP-CMT transport protocol |
| CN110769465B (en) * | 2019-10-24 | 2023-03-24 | Oppo（重庆）智能科技有限公司 | Data transmission method and device, electronic equipment and readable storage medium |
| CN111263402B (en) * | 2020-01-17 | 2023-04-07 | 广东中兴新支点技术有限公司 | Data transmission method and device and electronic equipment |
-
2022
- 2022-01-20 CN CN202210065202.9A patent/CN116527510A/en active Pending
- 2022-12-04 WO PCT/CN2022/136415 patent/WO2023138233A1/en unknown
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023138233A1 (en) | 2023-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106358245B (en) | Method and controller for sharing load of mobile edge computing application | |
| CN112769897B (en) | Synchronization method and device of edge calculation message, electronic equipment and storage medium | |
| CN110058937B (en) | Method, apparatus and medium for scheduling dedicated processing resources | |
| CN114466226B (en) | Bandwidth duration duty cycle determination method, device, equipment and computer readable medium | |
| CN112152879B (en) | Network quality determination method, device, electronic equipment and readable storage medium | |
| CN116527510A (en) | Model transmission method, device, electronic equipment and readable storage medium | |
| CN111245878A (en) | Method for computing and offloading communication network based on hybrid cloud computing and fog computing | |
| CN115567602A (en) | CDN node back-to-source method, device and computer readable storage medium | |
| CN114692898A (en) | MEC federal learning method, device and computer readable storage medium | |
| CN114253710A (en) | Processing method of computing request, intelligent terminal, cloud server, equipment and medium | |
| CN113485842A (en) | Method and device for analyzing data based on device cluster | |
| CN116527498A (en) | Model transmission method, device, electronic equipment and storage medium | |
| US11968122B2 (en) | Joint optimization method and system for delay and spectrum occupation in cloud-edge collaborative network | |
| CN116614367A (en) | Grouping method and model updating method of terminal equipment | |
| CN116614383A (en) | Model updating method and device, electronic equipment and storage medium | |
| US20230421501A1 (en) | Joint optimization method and system for delay and spectrum occupation in cloud-edge collaborative network | |
| CN116760785B (en) | Bandwidth allocation method, device, electronic equipment and storage medium | |
| CN117221127A (en) | Method and device for using artificial intelligence service at communication network end side | |
| CN116546586A (en) | Data transmission method, system, equipment and medium based on network physical layer security | |
| CN215729726U (en) | 5G edge computing platform architecture | |
| CN117235513A (en) | Model training method and model application method | |
| CN115086300B (en) | Video file scheduling method and device | |
| CN113568936B (en) | Real-time stream data storage method, device and terminal equipment | |
| CN113132269B (en) | System, method and computer readable storage medium for resource slice management | |
| WO2023138234A1 (en) | Model management method and apparatus, networking architecture, electronic device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |