onnx / tensorflow-onnx and FINN framework

[amroybd]

Hi, do the FINN framework support the ONNX format converted from the TFLite model using this tf2onnx converter?
Have a nice day!

[heborras]

Hi @amroybd,
currently FINN only supports quantized models, which were trained and exported by Brevitas.
In the future, we are planning to add support for models trained by QKeras, in collaboration with our colleagues at hls4ml.
Using the tf2onnx converter as it is currently will likely result in a network, which uses a different quantization to what FINN would expect from Brevitas or QONNX.

[amroybd]

I have used hls4ml. It was very interesting. Thank you for your reply