Initialize ORT in a separate isolate

lib/services/semantic_search/frameworks/onnx/onnx.dart

@@ -35,6 +35,7 @@ class ONNX extends MLFramework {
   @override
   Future<void> loadImageModel(String path) async {
     final startTime = DateTime.now();
+    await _clipImage.init();
     _imageEncoderAddress = await _computer.compute(
       _clipImage.loadModel,
       param: {
@@ -50,7 +51,9 @@ class ONNX extends MLFramework {
   @override
   Future<void> loadTextModel(String path) async {
     final startTime = DateTime.now();
-    await _clipText.init();
+    await _computer.compute(_clipText.init);
+    // Doing this from main isolate since `rootBundle` cannot be accessed outside it
+    await _clipText.initTokenizer();
     _textEncoderAddress = await _computer.compute(
       _clipText.loadModel,
       param: {

lib/services/semantic_search/frameworks/onnx/onnx_image_encoder.dart

@@ -9,7 +9,7 @@ import "package:onnxruntime/onnxruntime.dart";
 class OnnxImageEncoder {
   final _logger = Logger("OnnxImageEncoder");
 
-  OnnxImageEncoder() {
+  Future<void> init() async {
     OrtEnv.instance.init();
   }
 
@@ -96,14 +96,14 @@ class OnnxImageEncoder {
     final int ny = rgb.height;
     final int inputSize = 3 * nx * ny;
     final inputImage = List.filled(inputSize, 0.toDouble());
-  
+
     const int nx2 = 224;
     const int ny2 = 224;
     const int totalSize = 3 * nx2 * ny2;
 
     // Load image into List<double> inputImage
     for (int y = 0; y < ny; y++) {
-      for (int x = 0; x < nx; x ++) {
+      for (int x = 0; x < nx; x++) {
         final int i = 3 * (y * nx + x);
         inputImage[i] = rgb.getPixel(x, y).r.toDouble();
         inputImage[i + 1] = rgb.getPixel(x, y).g.toDouble();
@@ -121,7 +121,7 @@ class OnnxImageEncoder {
     final std = [0.26862954, 0.26130258, 0.27577711];
 
     for (int y = 0; y < ny3; y++) {
-      for (int x = 0; x < nx3; x ++) {
+      for (int x = 0; x < nx3; x++) {
         for (int c = 0; c < 3; c++) {
           //linear interpolation
           final double sx = (x + 0.5) * scale - 0.5;
@@ -152,18 +152,17 @@ class OnnxImageEncoder {
           final double v = v0 * (1 - dy) + v1 * dy;
 
           final int v2 = min(max(v.round(), 0), 255);
-          
+
           final int i = 3 * (y * nx3 + x) + c;
 
           result[i] = ((v2 / 255) - mean[c]) / std[c];
-
         }
       }
     }
     final floatList = Float32List.fromList(result);
 
-
-    final inputOrt = OrtValueTensor.createTensorWithDataList(floatList, [1, 3, 224, 224]);
+    final inputOrt =
+        OrtValueTensor.createTensorWithDataList(floatList, [1, 3, 224, 224]);
     final inputs = {'input': inputOrt};
     final session = OrtSession.fromAddress(args["address"]);
     final outputs = session.run(runOptions, inputs);

lib/services/semantic_search/frameworks/onnx/onnx_text_encoder.dart

@@ -8,19 +8,17 @@ import "package:onnxruntime/onnxruntime.dart";
 import "package:photos/services/semantic_search/frameworks/onnx/onnx_text_tokenizer.dart";
 
 class OnnxTextEncoder {
-  static const vocabFilePath = "assets/models/clip/bpe_simple_vocab_16e6.txt";
+  static const kVocabFilePath = "assets/models/clip/bpe_simple_vocab_16e6.txt";
   final _logger = Logger("OnnxTextEncoder");
   final OnnxTextTokenizer _tokenizer = OnnxTextTokenizer();
 
-  OnnxTextEncoder() {
+  Future<void> init() async {
     OrtEnv.instance.init();
-    OrtEnv.instance.availableProviders().forEach((element) {
-      _logger.info('onnx provider=$element');
-    });
   }
 
-  Future<void> init() async {
-    final vocab = await rootBundle.loadString(vocabFilePath);
+  // Do not run in an isolate since rootBundle can only be accessed in the main isolate
+  Future<void> initTokenizer() async {
+    final vocab = await rootBundle.loadString(kVocabFilePath);
     await _tokenizer.init(vocab);
   }