Cannot convert ONNX/TFLITE model to MNN

[K-prog]

Using the latest MNNConvert, I am not able to convert my original TFLITE model to MNN.

I tried converting it to ONNX and then exporting but it seems to get stuck in a loop and throws warnings, tried using onnx-sim and static batch sizes as well

Command used:

MNNConvert.exe -f ONNX --modelFile "facenet_dynamic-sim.onnx" --MNNModel facenet.mnn --bizCode MNN --info

Model Links

[jxt1234]

Use testMNNFromOnnx.py to test. The error may be caused by subgraph's inputs not valid.

[K-prog]

Seems like it stopped at:

Start to Optimize the MNN Net...

This didn't throw any error, also the ONNX model runs properly.

[jxt1234]

It seems has error with dequant. Could you use float model convert to MNN and then quantized it by MNN ?

[github-actions[bot]]

Marking as stale. No activity in 60 days.