Fixed the actual neural_net browser file that neede fixing.

Author: ParthPatel2000

src/game_2048/neural_net.js

@@ -1,147 +1,4 @@
 
-// // Complete ONNX setup that works in both Node.js and browser
-// let ort;
-// let session;
-// const MOVE_DIRECTIONS = ['up', 'down', 'left', 'right'];
-
-
-
-// function getModelPath() {
-//     if (typeof window !== 'undefined') {
-//         // Base-aware path
-//         return import.meta.env.BASE_URL + 'Onnx_models/model_2048_for_js.onnx';
-//     } else {
-//         // Node.js - can use relative path
-//         return 'public/Onnx_models/model_2048_for_js.onnx';
-//     }
-// }
-
-
-// async function initializeOrt() {
-//   if (!ort) {
-//     if (typeof window !== 'undefined') {
-//       // Browser
-//       const ortWeb = await import('onnxruntime-web');
-
-//     //   ortWeb.env.wasm.wasmPaths = 'https://cdn.jsdelivr.net/npm/onnxruntime-web@1.16.3/dist/';
-//       ortWeb.env.wasm.wasmPaths = import.meta.env.BASE_URL + 'assets/';
-//       ortWeb.env.wasm.numThreads = 1;
-//       ortWeb.env.logLevel = 'warning';
-
-//       try {
-//         await ortWeb.env.wasm.init();
-//       } catch (err) {
-//         console.warn('WASM init failed, fallback to CPU:', err);
-//       }
-
-//       ort = ortWeb;
-//     } else {
-//       // Node
-//       const ortNode = await import('onnxruntime-node');
-//       ort = ortNode;
-//     }
-//   }
-//   console.log('ONNX Runtime initialized:', ort);
-//   console.log("using " + (typeof window !== 'undefined' ? "onnxruntime-web" : "onnxruntime-node"));
-//   return ort;
-// }
-
-// async function loadModel(modelPath = getModelPath()) {
-//     await initializeOrt();
-
-//     try {
-//         const options = {};
-
-//         // For browser, specify execution providers explicitly
-//         if (typeof window !== 'undefined') {
-//             options.executionProviders = ['cpu']; // Safe fallback
-//         }
-
-//         session = await ort.InferenceSession.create(modelPath, options);
-//         console.log('Model loaded successfully');
-//     } catch (error) {
-//         console.error('Failed to load model:', error);
-
-//         // Try with different execution providers as fallback
-//         if (typeof window !== 'undefined') {
-//             try {
-//                 console.log('Trying with WASM execution provider...');
-//                 const wasmOptions = { executionProviders: ['wasm'] };
-//                 session = await ort.InferenceSession.create(modelPath, wasmOptions);
-//                 console.log('Model loaded with WASM provider');
-//             } catch (wasmError) {
-//                 console.error('WASM fallback also failed:', wasmError);
-//                 throw error;
-//             }
-//         } else {
-//             throw error;
-//         }
-//     }
-// }
-
-// export async function NeuralNet(gameboard) {
-//     console.log("in NeuralNet function");
-//     const preProcessBoard = (board) => {
-//         return board.flat().map(x => Math.log2(x === 0 ? 1 : x) / 16); //normalize by max tile (65536)
-//     };
-
-//     // Ensure model is loaded
-//     if (!session) {
-//         await loadModel();
-//     }
-
-//     // Ensure ort is initialized
-//     if (!ort) {
-//         await initializeOrt();
-//     }
-
-//     try {
-//         // Create input tensor
-//         const processedBoard = preProcessBoard(gameboard);
-//         console.log("Processed Board for NN:", processedBoard);
-
-//         const inputTensor = new ort.Tensor('float32',
-//             Float32Array.from(processedBoard),
-//             [1, processedBoard.length]
-//         );
-
-//         const feeds = { input: inputTensor };
-
-//         // Run inference
-//         const results = await session.run(feeds);
-//         const output = results.output.data; // Array of scores for each move
-
-//         // Find index of max score
-//         let maxIndex = 0;
-//         let maxScore = output[0];
-
-//         for (let i = 1; i < output.length; i++) {
-//             if (output[i] > maxScore) {
-//                 maxIndex = i;
-//                 maxScore = output[i];
-//             }
-//         }
-
-//         return {
-//             direction: MOVE_DIRECTIONS[maxIndex],
-//             confidence: maxScore,
-//             allScores: Array.from(output)
-//         };
-
-//     } catch (error) {
-//         console.error('Error during neural network inference:', error);
-//         throw error;
-//     }
-// }
-
-// // Optional: Add a function to cleanup resources
-// export function cleanup() {
-//     if (session) {
-//         session.release();
-//         session = null;
-//     }
-// }
-
 // Complete ONNX setup that works in both Node.js and browser
 let ort;
 let session;
@@ -150,7 +7,7 @@ const MOVE_DIRECTIONS = ['up', 'down', 'left', 'right'];
 // Base-aware model path
 function getModelPath() {
   if (typeof window !== 'undefined') {
-    return import.meta.env.BASE_URL + 'Onnx_models/model_2048_for_js.onnx';
+    return import.meta.env.BASE_URL + 'react_playground/Onnx_models/model_2048_for_js.onnx';
   } else {
     return 'public/Onnx_models/model_2048_for_js.onnx';
   }

src/game_2048/neural_net_browser.js

@@ -6,7 +6,7 @@ const MOVE_DIRECTIONS = ['UP', 'DOWN', 'LEFT', 'RIGHT'];
 
 function getModelPath() {
     // Browser: model must be in public directory
-    return '/Onnx_models/model_2048_for_js.onnx';
+    return import.meta.env.BASE_URL + 'Onnx_models/model_2048_for_js.onnx';
 }
 
 let session;