CN112862662A - Method and equipment for distributed training of transform-xl language model - Google Patents
Method and equipment for distributed training of transform-xl language model Download PDFInfo
- Publication number
- CN112862662A CN112862662A CN202110264864.4A CN202110264864A CN112862662A CN 112862662 A CN112862662 A CN 112862662A CN 202110264864 A CN202110264864 A CN 202110264864A CN 112862662 A CN112862662 A CN 112862662A
- Authority
- CN
- China
- Prior art keywords
- training
- data
- sequence
- sub
- language model
- 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
- 238000012549 training Methods 0.000 title claims abstract description 117
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012163 sequencing technique Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000000638 solvent extraction Methods 0.000 claims abstract description 7
- 230000000903 blocking effect Effects 0.000 claims description 5
- 238000005070 sampling Methods 0.000 abstract description 7
- 102000003779 Dipeptidyl-peptidases and tripeptidyl-peptidases Human genes 0.000 description 7
- 108090000194 Dipeptidyl-peptidases and tripeptidyl-peptidases Proteins 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Machine Translation (AREA)
Abstract
The invention relates to a method and equipment for distributed training of a transform-xl language model, which are applied to a process of training the transform-xl language model by DPP, and the method comprises the following steps: acquiring text corpus data for training a tranformer xl language model; sequencing all the text corpus data according to a context sequence; partitioning the sequenced text corpus data into blocks according to the number of GPUs (graphics processing units) so as to divide the text corpus data into a plurality of subdata; distributing different sub-data to different GPUs, wherein the GPU trains the sequence of the distributed sub-data to be consistent with the context sequence; and sequentially training the sub data through the GPUs to realize training of the transform-xl language model. According to the scheme, the sampling method in the DDP is reconstructed, the reconstructed DDP is used for training the tranformer xl, the training speed of massive text corpora is accelerated, the efficiency problem is solved, and the historical information of the tranformer xl model is kept.
Description
Technical Field
The invention relates to the technical field of machine translation, in particular to a method and equipment for distributed training of a transform-xl language model.
Background
The training of the language model uses large-scale text corpora, and if only a single GPU (Graphics Processing Unit) is used for Processing, the speed is slow, so that multiple GPUs are often used for training together. Common training modes include DP (data parallel) and DDP (decentralized data parallel).
As shown in fig. 2A, the DP mode is to slice data of one batch, allocate the sliced data to a plurality of GPUs for calculation, and then synchronize each GPU parameter to one master GPU for parameter update; DDP adopts an all-reduce (protocol) mode, and distributes the number of data participles batch to each GPU according to a certain sampling mode; each GPU can update parameters; compared with the DP speed, the DDP speed is greatly improved; at present, a tranformer xl model generally adopts a DP mode to train multiple cards, and because the context of the tranformer xl model is related, each batch is required to be trained in sequence; the manner of DDP is disturbed, so that the tranformer xl model loses the historical information. The speed of a tranformer xl model for DP training massive text corpora is low; training using the DDP approach loses historical information for the tranformer xl model.
Thus, there is a need for a better solution to the problems of the prior art.
Disclosure of Invention
The invention provides a method and equipment for training a transformer-xl language model in a distributed mode, which can solve the technical problem that the transformer xl model loses historical information when a DDP (distributed data processing) mode is used for training in the prior art.
The technical scheme for solving the technical problems is as follows:
the embodiment of the invention provides a method for training a transform-xl language model in a distributed mode, which is applied to the process of training the transform-xl language model by DPP, and comprises the following steps:
acquiring text corpus data for training a tranformer xl language model;
sequencing all the text corpus data according to a context sequence;
partitioning the sequenced text corpus data into blocks according to the number of GPUs (graphics processing units) so as to divide the text corpus data into a plurality of subdata;
distributing different sub-data to different GPUs, wherein the GPU trains the sequence of the distributed sub-data to be consistent with the context sequence;
and sequentially training the sub data through the GPUs to realize training of the transform-xl language model.
In a specific embodiment, the allocating the different sub-data to the different GPUs includes:
sequencing the subdata according to the context sequence to generate a first sequence, and sequencing the GPUs according to the training sequence to generate a second sequence;
for each sub-data, determining a ranking of the sub-data in a first sequence; determining the GPU corresponding to the ranking in the second sequence; and distributing the sub data to the determined GPU.
In a specific embodiment, the "training the sub-data sequentially by each GPU to implement training of a transform-xl language model" includes:
sequentially training the sub data once through each GPU to realize training of one pair of transform-xl language models;
and finally training the transformer-xl language model by using a plurality of pairs of training wheels for the transformer-xl language model.
In a specific embodiment, in the training process of each pair of transform-xl language models, the iteration sequence of each subdata is the same.
In a specific embodiment, the same iteration sequence of each sub-data is realized by closing the parameter configuration of the shuffle of the DDP.
The embodiment of the invention also provides a device for distributed training of a transform-xl language model, which is applied to the process of training the transform-xl language model by DPP, and the device comprises:
the acquisition module is used for acquiring text corpus data for training a tranformer xl language model;
the sorting module is used for sorting all the text corpus data according to the context sequence;
the blocking module is used for blocking the sequenced text corpus data according to the number of the GPUs so as to divide the text corpus data into a plurality of subdata;
the distribution module is used for distributing different sub-data to different GPUs, and the GPU trains that the sequence of the distributed sub-data is consistent with the context sequence;
and the training module is used for training the subdata sequentially through the GPUs to realize the training of the transform-xl language model.
In a specific embodiment, the allocation module includes:
the generating module is used for sequencing the subdata according to the context sequence to generate a first sequence, and sequencing the GPUs according to the training sequence to generate a second sequence;
the processing module is used for determining the ranking of the subdata in the first sequence aiming at each subdata; determining the GPU corresponding to the ranking in the second sequence; and distributing the sub data to the determined GPU.
In a specific embodiment, the training module includes:
the first round of module is used for training the subdata once sequentially through the GPUs to realize the training of a transform-xl language model by one round;
and the multi-wheel module is used for realizing the final training of the transform-xl language model through the training of the multi-wheel transform-xl language model.
In a specific embodiment, in the training process of each pair of transform-xl language models, the iteration sequence of each subdata is the same.
In a specific embodiment, the same iteration sequence of each sub-data is realized by closing the parameter configuration of the shuffle of the DDP
The invention has the beneficial effects that:
the embodiment of the invention provides a method and equipment for training a transform-xl language model in a distributed mode, which are applied to the process of training the transform-xl language model by DPP, and the method comprises the following steps: acquiring text corpus data for training a tranformer xl language model; sequencing all the text corpus data according to a context sequence; partitioning the sequenced text corpus data into blocks according to the number of GPUs (graphics processing units) so as to divide the text corpus data into a plurality of subdata; distributing different sub-data to different GPUs, wherein the GPU trains the sequence of the distributed sub-data to be consistent with the context sequence; and sequentially training the sub data through the GPUs to realize training of the transform-xl language model. According to the scheme, the sampling method in the DDP is reconstructed, the reconstructed DDP is used for training the tranformer xl, the training speed of massive text corpora is accelerated, the efficiency problem is solved, and the historical information of the tranformer xl model is kept.
Drawings
FIG. 1 is a flowchart illustrating a method for distributed training of a transform-xl language model according to an embodiment of the present invention;
FIG. 2A is a diagram illustrating data sampling in the prior art;
FIG. 2B is a schematic diagram of data sampling of a distributed method for training a transform-xl language model according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an apparatus for distributed training of a transform-xl language model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a partitioning module in an apparatus for distributed training of a transformer-xl language model according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a training module in a device for distributed training of a transformer-xl language model according to an embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
Example 1
The embodiment 1 of the invention discloses a method for training a transform-xl language model in a distributed manner, which is applied to a process of training the transform-xl language model by DPP (dipeptidyl peptidase), and as shown in FIG. 1, the method comprises the following steps:
101, acquiring text corpus data for training a tranformer xl language model;
first, text corpus data needs to be acquired for subsequent sampling.
as shown in FIG. 2, the text corpus data is sorted in context order as shown by 1. cndot. N in FIG. 2.
103, partitioning the sequenced text corpus data into blocks according to the number of GPUs (graphics processing units) so as to divide the text corpus data into a plurality of subdata;
further, as shown in fig. 2B, there are 3 GPUs, which are GPU1, GPU2, and GPU 3; the sorted text corpus is also divided into 3 blocks, namely three blocks 1 · 1/3 · N, 1/3 · N · 2/3 · N, 2/3 · N.
specifically, in step 104, the allocating the different sub-data to the different GPUs includes:
sequencing the subdata according to the context sequence to generate a first sequence, and sequencing the GPUs according to the training sequence to generate a second sequence;
for each sub-data, determining a ranking of the sub-data in a first sequence; determining the GPU corresponding to the ranking in the second sequence; and distributing the sub data to the determined GPU.
Specifically, as shown in fig. 2, the training sequence includes GPU1, GPU2, and GPU3, in this order, so that 1 · 1/3 · N is assigned to GPU1, 1/3 · N · 2/3 · N is assigned to GPU2, and 2/3 · N is assigned to GPU 3.
And 105, training the sub data sequentially through the GPUs to train a transform-xl language model.
Specifically, the "training the sub-data sequentially by each GPU to implement training of the transform-xl language model" in step 105 includes:
sequentially training the sub data once through each GPU to realize training of one pair of transform-xl language models;
and finally training the transformer-xl language model by using a plurality of pairs of training wheels for the transformer-xl language model.
Specifically, taking fig. 2B as an example for explanation, the GPU1, the GPU2, and the GPU3 are executed in sequence, so as to train sub data allocated to the GPU1, the GPU2, and the GPU3 in sequence, and complete a round of training process. And the complete training comprises multiple rounds of training, and the iteration sequence of each subdata in each round of training is the same.
Specifically, in the training process of each pair of transform-xl language models, the iteration sequence of each subdata is the same.
Further, the same iteration sequence of each subdata is realized by closing the parameter configuration of the shuffle of the DDP.
Example 2
The embodiment 2 of the present invention further discloses a distributed apparatus for training a transform-xl language model, which is applied to a process of training a transform-xl language model by DPP, and as shown in fig. 3, the apparatus includes:
an obtaining module 201, configured to obtain text corpus data for training a tranformer xl language model;
a sorting module 202, configured to sort all the text corpus data according to a context sequence;
a blocking module 203, configured to block the sequenced text corpus data according to the number of GPUs, so as to divide the text corpus data into a plurality of sub-data;
an allocating module 204, configured to allocate different sub-data to different GPUs, where an order of the allocated sub-data is trained by the GPU to be consistent with the context order;
the training module 205 is configured to train the sub-data sequentially through each GPU, so as to train a transform-xl language model.
In a specific embodiment, as shown in fig. 4, the allocating module 203 includes:
a generating module 2031, configured to rank the sub-data according to a context order to generate a first sequence, and rank the GPUs according to a training order to generate a second sequence;
a processing module 2032, configured to determine, for each of the sub-data, a ranking of the sub-data in the first sequence; determining the GPU corresponding to the ranking in the second sequence; and distributing the sub data to the determined GPU.
In a specific embodiment, as shown in fig. 5, the training module 205 includes:
the round module 2051 is used for training the sub data once sequentially through the GPUs to realize the training of a one-wheel transform-xl language model;
and a multi-round module 2052, configured to implement final training on the transform-xl language model by training the transform-xl language model through multiple rounds.
In a specific embodiment, in the training process of each pair of transform-xl language models, the iteration sequence of each subdata is the same.
In a specific embodiment, the same iteration sequence of each sub-data is realized by closing the parameter configuration of the shuffle of the DDP
The embodiment of the invention provides a method and equipment for training a transform-xl language model in a distributed mode, which are applied to the process of training the transform-xl language model by DPP, and the method comprises the following steps: acquiring text corpus data for training a tranformer xl language model; sequencing all the text corpus data according to a context sequence; partitioning the sequenced text corpus data into blocks according to the number of GPUs (graphics processing units) so as to divide the text corpus data into a plurality of subdata; distributing different sub-data to different GPUs, wherein the GPU trains the sequence of the distributed sub-data to be consistent with the context sequence; and sequentially training the sub data through the GPUs to realize training of the transform-xl language model. According to the scheme, the sampling method in the DDP is reconstructed, the reconstructed DDP is used for training the tranformer xl, the training speed of massive text corpora is accelerated, the efficiency problem is solved, and the historical information of the tranformer xl model is kept.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method for distributed training of a transform-xl language model, applied to a process for training a transform-xl language model by DPP, the method comprising:
acquiring text corpus data for training a tranformer xl language model;
sequencing all the text corpus data according to a context sequence;
partitioning the sequenced text corpus data into blocks according to the number of GPUs (graphics processing units) so as to divide the text corpus data into a plurality of subdata;
distributing different sub-data to different GPUs, wherein the GPU trains the sequence of the distributed sub-data to be consistent with the context sequence;
and sequentially training the sub data through the GPUs to realize training of the transform-xl language model.
2. The method of claim 1, wherein the assigning the different sub-data to the different GPUs comprises:
sequencing the subdata according to the context sequence to generate a first sequence, and sequencing the GPUs according to the training sequence to generate a second sequence;
for each sub-data, determining a ranking of the sub-data in a first sequence; determining the GPU corresponding to the ranking in the second sequence; and distributing the sub data to the determined GPU.
3. The method of claim 1, wherein the "training the sub-data to achieve training of the transform-xl language model sequentially through the GPUs" comprises:
sequentially training the sub data once through each GPU to realize training of one pair of transform-xl language models;
and finally training the transformer-xl language model by using a plurality of pairs of training wheels for the transformer-xl language model.
4. The method of claim 3, wherein the order of iteration of each of the sub-data is the same during training of each pair of transform-xl language models.
5. The method of claim 4, wherein the sub-data is iterated in the same order by closing a parameter configuration of a shuffle of the DDP.
6. An apparatus for distributed training of a transform-xl language model, applied to a process of training the transform-xl language model by DPP, the apparatus comprising:
the acquisition module is used for acquiring text corpus data for training a tranformer xl language model;
the sorting module is used for sorting all the text corpus data according to the context sequence;
the blocking module is used for blocking the sequenced text corpus data according to the number of the GPUs so as to divide the text corpus data into a plurality of subdata;
the distribution module is used for distributing different sub-data to different GPUs, and the GPU trains that the sequence of the distributed sub-data is consistent with the context sequence;
and the training module is used for training the subdata sequentially through the GPUs to realize the training of the transform-xl language model.
7. The apparatus of claim 6, wherein the assignment module comprises:
the generating module is used for sequencing the subdata according to the context sequence to generate a first sequence, and sequencing the GPUs according to the training sequence to generate a second sequence;
the processing module is used for determining the ranking of the subdata in the first sequence aiming at each subdata; determining the GPU corresponding to the ranking in the second sequence; and distributing the sub data to the determined GPU.
8. The apparatus of claim 6, wherein the training module comprises:
the first round of module is used for training the subdata once sequentially through the GPUs to realize the training of a transform-xl language model by one round;
and the multi-wheel module is used for realizing the final training of the transform-xl language model through the training of the multi-wheel transform-xl language model.
9. The apparatus of claim 8, wherein the order of iteration of each of the sub-data is the same during training of each pair of transform-xl language models.
10. The apparatus of claim 9, wherein the sub-data is iterated in the same order by closing a parameter configuration of a shuffle of the DDP.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110264864.4A CN112862662A (en) | 2021-03-12 | 2021-03-12 | Method and equipment for distributed training of transform-xl language model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110264864.4A CN112862662A (en) | 2021-03-12 | 2021-03-12 | Method and equipment for distributed training of transform-xl language model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112862662A true CN112862662A (en) | 2021-05-28 |
Family
ID=75994045
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110264864.4A Pending CN112862662A (en) | 2021-03-12 | 2021-03-12 | Method and equipment for distributed training of transform-xl language model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112862662A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115589446A (en) * | 2022-09-26 | 2023-01-10 | 黑盒科技(广州)有限公司 | Meeting abstract generation method and system based on pre-training and prompting |
WO2023241312A1 (en) * | 2022-06-16 | 2023-12-21 | 北京火山引擎科技有限公司 | Model training method and apparatus |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102509549A (en) * | 2011-09-28 | 2012-06-20 | 盛乐信息技术(上海)有限公司 | Language model training method and system |
CN105786782A (en) * | 2016-03-25 | 2016-07-20 | 北京搜狗科技发展有限公司 | Word vector training method and device |
US20190156215A1 (en) * | 2017-11-18 | 2019-05-23 | Neuralmagic Inc. | Systems and methods for exchange of data in distributed training of machine learning algorithms |
CN110379416A (en) * | 2019-08-15 | 2019-10-25 | 腾讯科技(深圳)有限公司 | A kind of neural network language model training method, device, equipment and storage medium |
CN110503194A (en) * | 2019-08-09 | 2019-11-26 | 苏州浪潮智能科技有限公司 | A kind of method and system of distributed parallel training |
CN110705705A (en) * | 2019-09-25 | 2020-01-17 | 浪潮电子信息产业股份有限公司 | Convolutional neural network model synchronous training method, cluster and readable storage medium |
CN111159416A (en) * | 2020-04-02 | 2020-05-15 | 腾讯科技(深圳)有限公司 | Language task model training method and device, electronic equipment and storage medium |
CN111381966A (en) * | 2020-03-08 | 2020-07-07 | 苏州浪潮智能科技有限公司 | Distributed parallel training method, device and readable medium |
CN111832292A (en) * | 2020-06-03 | 2020-10-27 | 北京百度网讯科技有限公司 | Text recognition processing method and device, electronic equipment and storage medium |
-
2021
- 2021-03-12 CN CN202110264864.4A patent/CN112862662A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102509549A (en) * | 2011-09-28 | 2012-06-20 | 盛乐信息技术(上海)有限公司 | Language model training method and system |
CN105786782A (en) * | 2016-03-25 | 2016-07-20 | 北京搜狗科技发展有限公司 | Word vector training method and device |
US20190156215A1 (en) * | 2017-11-18 | 2019-05-23 | Neuralmagic Inc. | Systems and methods for exchange of data in distributed training of machine learning algorithms |
CN110503194A (en) * | 2019-08-09 | 2019-11-26 | 苏州浪潮智能科技有限公司 | A kind of method and system of distributed parallel training |
CN110379416A (en) * | 2019-08-15 | 2019-10-25 | 腾讯科技(深圳)有限公司 | A kind of neural network language model training method, device, equipment and storage medium |
CN110705705A (en) * | 2019-09-25 | 2020-01-17 | 浪潮电子信息产业股份有限公司 | Convolutional neural network model synchronous training method, cluster and readable storage medium |
CN111381966A (en) * | 2020-03-08 | 2020-07-07 | 苏州浪潮智能科技有限公司 | Distributed parallel training method, device and readable medium |
CN111159416A (en) * | 2020-04-02 | 2020-05-15 | 腾讯科技(深圳)有限公司 | Language task model training method and device, electronic equipment and storage medium |
CN111832292A (en) * | 2020-06-03 | 2020-10-27 | 北京百度网讯科技有限公司 | Text recognition processing method and device, electronic equipment and storage medium |
Non-Patent Citations (2)
Title |
---|
LITTLEWHITE: ""最全PyTorch分布式教程"来了!", pages 1 - 7, Retrieved from the Internet <URL:https://zhuanlan.zhihu.com/p/267157806> * |
朱祥磊;: "加速AI分布式训练研究和实践", 电信技术, no. 12, 25 December 2019 (2019-12-25), pages 28 - 31 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023241312A1 (en) * | 2022-06-16 | 2023-12-21 | 北京火山引擎科技有限公司 | Model training method and apparatus |
CN115589446A (en) * | 2022-09-26 | 2023-01-10 | 黑盒科技(广州)有限公司 | Meeting abstract generation method and system based on pre-training and prompting |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112862662A (en) | Method and equipment for distributed training of transform-xl language model | |
CN111553484A (en) | Method, device and system for federal learning | |
DE102017008956A1 (en) | Method for using a computer unit | |
CN104408192B (en) | The compression processing method and device of character string type row | |
CN115204413A (en) | Intelligent learning data processing method based on artificial intelligence | |
CN109948632A (en) | Data training method, apparatus and system, computer equipment | |
CN106548674A (en) | For the method for dynamically processing and system of net marking | |
CN117575044A (en) | Data forgetting learning method, device, data processing system and storage medium | |
CN112199885B (en) | Distributed model training system and application method | |
WO2001018642A3 (en) | Method and apparatus for synchronizing function values in a multiple protocol system | |
CN111695689B (en) | Natural language processing method, device, equipment and readable storage medium | |
CN113326329A (en) | Method for pushing account-keeping person based on block chain consensus | |
CN111695701B (en) | System for realizing data set construction processing based on federal learning and construction generation method thereof | |
CN110189465B (en) | Random lottery method and equipment by means of block chain | |
CN111949786A (en) | Intelligent question-answer model optimization method and device | |
CN106203632A (en) | A kind of limited knowledge collection recombinant is also distributed the study of extraction and application system method | |
CN104750560B (en) | A kind of information processing method and electronic equipment | |
CN103714591B (en) | Rail vehicle operation data storage method and data recording equipment | |
CN110825453B (en) | Data processing method and device based on big data platform | |
EP4325357A3 (en) | Clustering processes using traffic data | |
CN103473374B (en) | Patient data partitioning system and patient data partitioning method | |
CN107133639A (en) | Merge the adaptively sampled method of non-equilibrium data of Boost models | |
CN106354581A (en) | Cyclic redundancy check method and multi-core processor | |
CN109548153B (en) | Resource allocation method based on modularity SC-FDMA system | |
CN112861549A (en) | Method and equipment for training translation model |
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 |