@mineproness/tensorloop 1.2.9 → 1.3.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mineproness/tensorloop",
-  "version": "1.2.9",
+  "version": "1.3.0",
   "description": "The most powerfull Machine Learning Library for AI.",
   "keywords": [
     "ai",

package/src/Linear/index.js

@@ -1,4 +1,4 @@
-import { LoadBias,LoadVectors,SaveBias,SaveVectors } from '../GetConfigs.js'
+import { LoadBias, LoadVectors, SaveBias, SaveVectors } from '../GetConfigs.js'
 import { existsSync } from 'fs'
 function GenerateWeightsBias(embeddingSize, vocabSize) {
     const vectors = []
@@ -30,8 +30,8 @@ export class Linear {
                 const { vectors, Bias } = GenerateWeightsBias(embeddingSize, vocabSize)
                 this.Weights = vectors;
                 this.Bias = Bias;
-                SaveVectors( this.Weights,configs.save.filename[0])
-                SaveBias(this.Bias,configs.save.filename[1])
+                SaveVectors(this.Weights, configs.save.filename[0])
+                SaveBias(this.Bias, configs.save.filename[1])
             }
         } else {
             const { vectors, Bias } = GenerateWeightsBias(embeddingSize, vocabSize)
@@ -40,7 +40,7 @@ export class Linear {
         }
     }
     Save() {
-        SaveVectors(this.Weights,this.configs.save.filename[0])
+        SaveVectors(this.Weights, this.configs.save.filename[0])
         SaveBias(this.Bias, this.configs.save.filename[1])
     }
     forward(input) {
@@ -55,19 +55,19 @@ export class Linear {
         }
         return output
     }
-    backward(outputGradient, learningRate, idx) {
-        const input = this.input[idx];
-        const weights = this.Weights
-        const inputGradient = new Float32Array(input.length)
-        for (let a = 0; a < weights.length; a++) {
-            const gran = outputGradient[a]
-            const w = weights[a]
-            for (let at = 0; at < input.length; at++) {
-                inputGradient[at] += gran * w[at]
-                w[at] -= learningRate * input[at] * gran
+    backward(dOut, lr, stepIndex) {
+        let dIn = new Float32Array(this.input[stepIndex]);
+        for (let i = 0; i < this.inputDim; i++) {
+            let gradSum = 0;
+            for (let j = 0; j < this.outputDim; j++) {
+                gradSum += dOut[j] * this.weights[i][j];
+                this.weights[i][j] -= lr * dOut[j] * this.input[stepIndex][i];
             }
-            this.Bias[a] -= learningRate * gran
+            dIn[i] = gradSum;
         }
-        return inputGradient
+        for (let j = 0; j < this.outputDim; j++) {
+            this.bias[j] -= lr * dOut[j];
+        }
+        return dIn;
     }
 }