How you debug/measure Python -> Javascript ONNX Conversion

[matbee-eth]

Question

I have converted a couple ONNX models to use ONNXRuntimeWeb from using the Python onnx version as the source. Ive spent weeks debugging though. What's your strategy for comparing tensor values, etc, with these onnx models?

Ive console log'd N# of values from the tensor/array to see if the values have diverged far but it can get fatiguing. I can't simply just dump a numpy array and compare