Android-oriented deep learning model unified deployment system, method, device and medium
A technology for deep learning and deployment systems, applied in the field of unified deployment systems for deep learning models, which can solve the problems of lack of automatic resource management, complicated deployment operation APIs, and low level of runtime integration.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0050] like figure 1 Shown:
[0051]The present disclosure provides an Android-oriented deep learning model unified deployment system,
[0052] include:
[0053] The monitoring and statistics module is used to complete the monitoring and statistics of the entire system and log reading and writing;
[0054] The abstract adaptation module is used to provide the interpreter, model and / or data source required by the deep learning inference process;
[0055] The service module is used to provide a unified programming interface for Android applications.
[0056] Specifically, the unified deployment framework is mainly composed of three modules:
[0057] Abstract Adaptation Module: Provides three abstractions of interpreter, model, and data source required for deep learning inference tasks. For different underlying deep learning frameworks, including TensorFlow Lite, TVM, MNN, etc., adapt a corresponding specific interpreter singleton; for different model files, create model ins...
Embodiment 2
[0083] like figure 2 as shown,
[0084] The present disclosure can also provide a unified deployment method for Android-oriented deep learning models, which is applied to the unified deployment system for Android-oriented deep learning models in the first embodiment above, including:
[0085] S201: Initialize the unified deployment system;
[0086] Specifically, the S201 specifically includes:
[0087] Load data sources and / or models, initialize interpreter context.
[0088] S202: Construct an inference task and execute the inference task to obtain an execution result;
[0089] Specifically, the S202 specifically includes:
[0090] Create an inference task, specifying the model, data source, and / or interpreter and specifying actions before and after the task starts;
[0091] Load the single input data, call the interpreter to interpret the model, and calculate the execution result of the inference task;
[0092] The execution result of the inference task is obtained.
...
Embodiment 3
[0105] The present disclosure can also provide a computer storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it is used to implement the above-mentioned steps of the Android-oriented deep learning model unified deployment method.
[0106] The computer storage medium of the present disclosure may be implemented using semiconductor memory, magnetic core memory, magnetic drum memory, or magnetic disk memory.
[0107] Semiconductor memory, mainly used in computers, mainly has two types of semiconductor memory elements: Mos and bipolar. Mos components are highly integrated, the process is simple but the speed is slow. Bipolar components are complex in process, high in power consumption, low in integration but fast in speed. After the advent of NMos and CMos, Mos memory began to play a major role in semiconductor memory. NMos is fast, for example, the access time of Intel's 1K-bit SRAM is 45ns. CMos consumes less power...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More