Muru

Muru is a machine learning backend system developed by CarnationStudio.

Usage

Muru is divided into four main components

• muru.lib
• muru.gpu
• muru.runtime
• muru.callback

muru.lib allows you to import and remove libraries related to Muru.
muru.gpu allows you to change the backend devices used by Muru.
muru.runtime allows EasyRun to easily run models.
muru.callback handles problems that occur in Muru and provides information as a CallBack.

Usage of muru.lib

muru.lib has one API.

muru.lib.import(lib)API

You can import any library that Muru supports.
The lib part is the name of the library as a String.
The available libraries are as follows.

• Legend: library name => import name
• Tensorflow.js => tfjs
• GPU.js => gpujs
• ml5.js => ml5.js
• Tensorflow-Model-Quna=> tfjsqna
  Example: Import Tensorflow.js library

import muru from 'Path-to-Muru';.
//import muru
muru.lib.import(“tfjs”);
// import Tensorflow.js here

Usage of muru.gpu

muru.gpu has three components.

muru.gpu.change()API

This API changes the backend device to the GPU.
Example: ``muru.gpu.change()

import muru from 'Path-to-Muru';
//import muru
muru.gpu.change()
// change backend device to GPU.

muru.gpu.cpu()API

This API changes the backend device to the CPU.
If you change the backend device to CPU, you will get only about 1/100 of the performance when using GPU.
Example:

import muru from 'Path-to-Muru';
//import muru
muru.gpu.cpu();
//set backend device to CPU

muru.gpu.wasmAPI

This API changes the backend device to WASM (WebAssembly).
The performance will be dramatically improved compared to CPU.
Example:

import muru from 'Path-to-Muru';
//import muru
muru.gpu.wasm();
//set backend device to wasm

Usage of muru.runtime

muru.runtime has one API.

muru.runtime.modelAPI

This API allows importing trained models.
Usage:

import muru from 'Path-to-Muru';
//import muru

muru.runtime.model.modelName.init();//init();//initialize model
muru.runtime.model.model name.use(input);//use model
/*Example: Qna model in Tensorflow.js (BERT Q&A model)*/
muru.runtime.model.qna.init();//initialize Qna
let result = muru.runtime.model.qna.use(q,a);//enter question a about q
console.log(result);//display output

muru.runtime.model.onnxAPI

This API allows you to run ONNX models.
Usage:

import muru from 'Path-to-Muru';
//import muru
(async () => {
   muru.runtime.onnx.init(modelpath);  
   //initialization ONNX model
   let result = await async muru.runtime.onnx.use(Input,Input2)
   //If you do not want Input2, enter a space instead of null.
   console.log(result);
   //Show Results
})();

Credits

©Carnation Studio 2022-2024

Muru BackEnd Runtime For JavaScript

Provided under the Mozilla Public License Version 2.0.

See the License file for the detailed license.