@tensorflow/tfjs初探

&＃64;tensorflow/tfjs 初探

• 一维张量&＃43;标量
• 二维张量&＃43;标量
• 三维张量&＃43;标量

名词标注翻译

• tensor 张量

  张量是向量或矩阵等更高维度的核心数据结构&＃xff0c;常用的有标量、一维、二维、三维和四维张量

• scalar 标量
• buffer 缓冲区

Tensor方法

• flatten()
• asScalar()
• as1D()
• as2D()
• as3D()
• as4D()
• asType()
• buffer()
• data()
• dataSync()
• dispose()
• toFloat()
• toInt()
• toBool()
• print(verbose: boolean)
• reshape(newShape)
• reshapeAs(x: Tensor)
• expandDims (axis?)
• squeeze (axis?)
• clone ()
• toString ()

• Model

• compile()
• evaluate()
• predict()
• predictOnBatch()
• fit()

• Sequential

• add(layer)
• evaluate (x, y, config?)
• predict (x, config?)
• fit (x, y, config?)

• Layer

• apply (inputs, kwargs?)

• Optimizer

• minimize(f, returnCost?, varList?)

封装成自己 service

• tf.service.ts

import { Injectable } from "&＃64;angular/core";
import {scalar,Scalar,Tensor,tensor,Tensor1D,tensor1d,Tensor2D,tensor2d,Tensor3D,tensor3d,Tensor4D,tensor4d,buffer,TensorBuffer,fill,fromPixels,linspace,oneHot,ones,onesLike,print,randomNormal,randomUniform,range,truncatedNormal,Rank,variable,Variable,zeros,zerosLike,cast,expandDims,pad,reshape,squeeze,concat,gather,reverse,slice,stack,tile,Sequential,sequential,SequentialConfig,model,Model,SymbolicTensor,input,loadModel
} from "&＃64;tensorflow/tfjs";
import { Subject } from "rxjs/Subject";
import { DType } from "&＃64;tensorflow/tfjs-layers/dist/types";
export declare type Shape &＃61; number[];
export interface InputConfig {shape?: Shape;batchShape?: Shape;name?: string;dtype?: DType;sparse?: boolean;
}
export interface ContainerConfig {inputs: SymbolicTensor | SymbolicTensor[];outputs: SymbolicTensor | SymbolicTensor[];name?: string;
}
export interface ShapeMap {R0: number[];R1: [number];R2: [number, number];R3: [number, number, number];R4: [number, number, number, number];
}
export interface DataTypeMap {float32: Float32Array;int32: Int32Array;bool: Uint8Array;
}
export declare type DataType &＃61; keyof DataTypeMap;/**
- tensor 张量
- scalar 标量
*/&＃64;Injectable()
export class TfService {constructor() {}tensor(values: any, shape?: number[], dtype?: DType): Tensor {return tensor(values, shape, dtype);}// 生成一维张量tensor1d(arr: any[], dtype?: DType): Tensor1D {return tensor1d(arr, dtype);}// 生成二维张量tensor2d(arr: any[][], shape?: [number, number], dtype?: DType): Tensor2D {return tensor2d(arr, shape, dtype);}// 生成三维张量tensor3d(arr: any[][][],shape?: [number, number, number],dtype?: DType): Tensor3D {return tensor3d(arr, shape, dtype);}// 生成四维张量tensor4d(arr: number[][][][],shape?: [number, number, number, number],dtype?: DType): Tensor4D {return tensor4d(arr, shape, dtype);}// 生成标量scalar(value: number | boolean,dtype?: "float32" | "int32" | "bool"): Scalar {return scalar(value);}// 缓冲区buffer(shape: number[], dtype?: DType, values?: any): TensorBuffer<any> {return buffer(shape, dtype, values);}// 复制一个 张量clone(x: Tensor): Tensor {return x.clone();}// 填充获取 张量fill(shape: number[], value: number, dtype?: DType): Tensor {return fill(shape, value, dtype);}// 通过图片创建一个 3D张量fromPixels(pixels: ImageData | HTMLImageElement | HTMLCanvasElement | HTMLVideoElement,numChannels?: number): Tensor3D {return fromPixels(pixels, numChannels);}// 制定区间返回一个均匀分布的一维张量linspace(start, stop, num): Tensor1D {return linspace(start, stop, num);}// 在制定位置放置制定数值 创建一个二维张量oneHot(indices: Tensor1D,depth: number,onValue?: number,offValue?: number): Tensor2D {return oneHot(indices, depth, onValue, offValue);}// 制定形状填充1ones(shape: number[], dtype?: DType): Tensor {return ones(shape, dtype);}// 创建一个与之相同形状的填充为1的张量onesLike(x: Tensor): Tensor {return onesLike(x);}// 打印数据print(x: Tensor, verbose?: boolean): void {return print(x, verbose);}// 根据参数随机填充张量randomNormal(shape: number[],mean?: number,stdDev?: number,dtype?: "float32" | "int32",seed?: number): Tensor {return randomNormal(shape, mean, stdDev, dtype, seed);}// 制定范围填充张量randomUniformextends Rank>(shape: ShapeMap[R],minval?: number,maxval?: number,dtype?: DataType): Tensor {return randomUniform(shape, minval, maxval, dtype);}// 在所提供的范围内填充数字range(start: number,stop: number,step?: number,dtype?: "float32" | "int32"): Tensor1D {return range(start, stop, step, dtype);}// 具有截断正态分布的采样值truncatedNormalextends Rank>(shape: ShapeMap[R],mean?: number,stdDev?: number,dtype?: "float32" | "int32",seed?: number): Tensor {return truncatedNormal(shape, mean, stdDev, dtype, seed);}// 用所提供的初始值创建一个新变量variableextends Rank>(initialValue: Tensor,trainable?: boolean,name?: string,dtype?: DataType): Variable {return variable(initialValue, trainable, name, dtype);}// 创建一个所有元素设置为0的张量zerosextends Rank>(shape: ShapeMap[R], dtype?: DataType): Tensor {return zeros(shape, dtype);}// 创建一个张量&＃xff0c;所有元素设置为0&＃xff0c;其形状与给定的张量相同zerosLikeextends Tensor>(x: T): T {return zerosLike(x);}// 投射到一个新的张量。castextends Tensor>(x: T, dtype: DataType): T {return cast(x, dtype);}// 通过在张量的形状中插入一个维度来扩展秩expandDimsextends Rank>(x: Tensor, axis?: number): Tensor {return expandDims(x, axis);}// 铺垫padextends Tensor>(x: T,paddings: Array<[number, number]>,constantValue?: number): T {return pad(x, paddings, constantValue);}// 改变形状reshapeextends Rank>(x: Tensor, shape: ShapeMap[R2]): Tensor {return reshape(x, shape);}// 消除尺寸squeezeextends Tensor>(x: Tensor, axis?: number[]): T {return squeeze(x, axis);}// 合并concatextends Tensor>(tensors: T[], axis?: number): T {return concat(tensors, axis);}// 根据指数收集切片gatherextends Tensor>(x: T, indices: Tensor1D, axis?: number): T {return gather(x, indices, axis);}// 反转reverseextends Tensor>(x: T, axis?: number | number[]): T {return reverse(x, axis);}// 分割sliceextends Rank, T extends Tensor>(x: T,begin: ShapeMap[R],size: ShapeMap[R]): T {return slice(x, begin, size);}// 堆砌stackextends Tensor>(tensors: T[], axis?: number): Tensor {return stack(tensors, axis);}// 重复tileextends Tensor>(x: T, reps: number[]): T {return tile(x, reps);}// sequential模型sequential(config?: SequentialConfig): Sequential {return sequential(config);}// 创建模型model(config: ContainerConfig): Model {return model(config);}// 输入input(config: InputConfig): SymbolicTensor {return input(config);}// 加载模型loadModel(modelConfigPath: string): Promise {return loadModel(modelConfigPath);}// 相加add(a: Tensor1D | Tensor2D | Tensor3D | Tensor4D, b: Scalar) {let sub: Subject<any> &＃61; new Subject();let result &＃61; a.add(b).data().then(res &＃61;> {sub.next(res);});return sub;}
}


实例

创建一个2行3列的矩阵

const a &＃61; this.tf.tensor([1.0, 2.0, 3.0, 10.0, 20.0, 30.0], [2, 3]);
a.print();
/**
Tensor
[[1 , 2 , 3 ],[10, 20, 30]
]
*/