fix: prediction issue resolved

src/app/services/chess-predictor.service.ts

@@ -1,5 +1,6 @@
 import { Injectable } from '@angular/core';
 import * as ort from 'onnxruntime-web';
+import { environment } from 'src/environments/environment';
 
 @Injectable({
   providedIn: 'root'
@@ -9,13 +10,28 @@ export class ChessPredictorService {
   private model: ort.InferenceSession | null = null;
 
   constructor() {
+    this.initializeBackend();
     this.loadModel();
   }
 
+  private async initializeBackend() {
+    try {
+      // Set the correct paths to the WASM files
+      ort.env.wasm.wasmPaths = '/assets/onnxruntime/';
+      console.log('WASM paths set to:', ort.env.wasm.wasmPaths);
+    } catch (err) {
+      console.error('Failed to set WASM paths:', err);
+    }
+  }
+
   // Load the ONNX model from the assets folder
   private async loadModel() {
     try {
-      this.model = await ort.InferenceSession.create('/assets/neural_network_model/chess_move_predictor.onnx');
+      const response = await fetch(`${environment.base_url}/nn_model/getNNModel`);
+      const arrayBuffer = await response.arrayBuffer();
+      const modelTensor = new Uint8Array(arrayBuffer);
+
+      this.model = await ort.InferenceSession.create(modelTensor);
       console.log('ONNX Model loaded successfully.');
     } catch (error) {
       console.error('Error loading ONNX model:', error);
@@ -24,6 +40,7 @@ export class ChessPredictorService {
 
   // Method to preprocess input data and make predictions
   public async predict(boardState: Float32Array): Promise<number | null> {
+    debugger;
     if (!this.model) {
       console.error('Model is not loaded yet.');
       return null;
@@ -35,10 +52,12 @@ export class ChessPredictorService {
     try {
       const feeds: ort.InferenceSession.OnnxValueMapType = { input: inputTensor };
       const results = await this.model.run(feeds);
-      const output = results['output'].data;
-
-      // Assert the type explicitly to number
-      const predictedMoveIndex = output[0] as number; // Treat output[0] as a number
+      const outputTensor = results['output'];  // Make sure this matches your ONNX model output name
+      const outputArray = Array.from(outputTensor.data as Float32Array);
+
+      // Find the index with the maximum value
+      const predictedMoveIndex = outputArray.indexOf(Math.max(...outputArray));
+
       return predictedMoveIndex;
     } catch (error) {
       console.error('Error during model inference:', error);