React Native: Can't load a model: Can't create InferenceSession

[Shahzaib-MST]

I have converted the Keras model to onnx using onnxmltools on python. the summary of the model is as below:
Model: "sequential"

---

Layer (type) Output Shape Param #

dense (Dense) (None, 32) 480

---

dense_1 (Dense) (None, 64) 2112

---

dense_2 (Dense) (None, 128) 8320

---

dense_3 (Dense) (None, 64) 8256

---

dropout (Dropout) (None, 64) 0

---

dense_4 (Dense) (None, 5) 325

Total params: 19,493
Trainable params: 19,493
Non-trainable params: 0

The converted model is working fine when used on python. In order to use the onnx model in react native, I converted the onnx model to ort format using the script:
python -m onnxruntime.tools.convert_onnx_models_to_ort model.onnx

The converted ORT model works perfectly fine on python but when I use this model in react native it throws exceptions while loading the model
the exceptions is:
[Error: Can't load a model: Can't create InferenceSession]

[hanbitmyths]

ONNX Runtime react native packages supports a limited set of operators in here. Can you check if all kernels in your model are supported and opset version is either 12 or 13?

If your model is not using opset 12/13, you can either export an ONNX model with opset 12/13 or you can convert onnx version using this script. Then, convert to ort.

[Shahzaib-MST]

Hi,
Yes, I have checked the required operators of my model are listed in the set of operators list

[stale]

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.