qonnx icon indicating copy to clipboard operation
qonnx copied to clipboard
verified publisher icon fastmachinelearning

Is tensor representable

Open jurevreca12 opened this issue 1 year ago • 0 comments

Adds a single function to the QKeras converter, to check if the tensor is being correctly converted. It issues only a warning to not brake legacy testing code.

This PR builds upon PR: Update qkeras converter for array and auto po2 scale #135 This is why it currently shows many differences, after they will be merged into main the only difference is in src/qonnx/converters/qkeras/onnx.py with the addition of the check_tensor_is_representable function.

This PR does not affect other code, nor should it be a problem for downstream users. It can be merged after PR: #135.

jurevreca12 avatar Aug 26 '24 08:08 jurevreca12