Description
Hello,
I have the following error when trying to load a very simple tf model in my react expo app:
Error loading model: Error: modelWeightsID must be a number or number array.
Have you wrapped your asset paths in a require() statements?
Here is the code snippet in my component.tsx:
import * as tf from "@tensorflow/tfjs";
import {bundleResourceIO, decodeJpeg} from "@tensorflow/tfjs-react-native";
import * as FileSystem from 'expo-file-system';
import * as DocumentPicker from 'expo-document-picker';
export default function App() {
const loadingtest = async ()=>{
console.log(loading);
await tf.ready();
console.log(tf ready);
const modelJson = await require('@/assets/model/newtest/model.json');
const weightAssets = await require('@/assets/model/newtest/group1-shard1of1.bin');
console.log(`modeljsobn`, modelJson);
console.log(`modelweigts`, weightAssets);
try {
const model = await tf.loadLayersModel(bundleResourceIO(modelJson, weightAssets));
console.log(`model loaded`);
// You can now use the 'model' object for predictions
} catch (error) {
console.error("Error loading model:", error);
}
}
loadingtest();
return (
<Animated.View style={[styles.container ]}>
</Animated.View>
);
}
System information
I created the model from GoogleColab, then installed Tensorflowjs to convert the model from .keras and .h5 to .json and .bin files for expo app.
The version for tensorflowconverter:
installed ml_dtypes-0.5.1 packaging-23.2 tensorboard-2.19.0 tensorflow-2.19.0 tensorflow-decision-forests-1.12.0 tensorflowjs-4.22.0 tf-keras-2.19.0 wurlitzer-3.1.1 ydf-0.11.0
- I tried on ios simulator, expo app go, and web browser chromem all unsuccessfull.
- I installed the following in my expo app from the page at https://www.npmjs.com/package/@tensorflow/tfjs-react-native:
expo-gl
expo-camera
async-storage
react-native-fs
@tensorflow/tfjs - npm install @tensorflow/tfjs
@tensorflow/tfjs-react-native - npm install @tensorflow/tfjs-react-native
Note: I had dependency conflict error and thus used the instruction --legacy-peer-deps
Describe the current behavior
in the console log on web browser, I get:
modeljsobn {format: 'layers-model', generatedBy: 'keras v3.8.0', convertedBy: 'TensorFlow.js Converter v4.22.0', modelTopology: {…}, weightsManifest: Array(1)}
rankview.tsx:29 modelweigts /assets/?unstable_path=.%2Fassets%2Fmodel%2Fnewtest/group1-shard1of1.bin
when doing console.log(modeljsobn, modelJson) and console.log(modelweigts, weightAssets); in the code snippet above.
then, the error: rankview.tsx:36 Error loading model: Error: modelWeightsID must be a number or number array.
Have you wrapped your asset paths in a require() statements?
Describe the expected behavior
the model should be loading as per the code snippet.
Standalone code to reproduce the issue
I am providing the two files .json and .bin to load the model in expo app as a zipped file archive.zip