Yaman Umuroglu comments

Results 47 comments of


                                            Yaman Umuroglu

`Resize` node empty `RoI` input cause issue for shape inference

I merged something that might fix for this as part of #125 , please give it a try and let me know.

Overhaul shape inference for custom ops

Thanks for sharing the results of your investigation @jurevreca12 ! Another emerging option is actually using functions in ONNX instead of the custom PyOp from onnxruntime-extensions. I'm planning to look...

Add resize to change_3d_tensors_to_4d transformation

Thanks @klassen9 . Since this is also related to #150 I took the liberty to extend the testcase generation to add a opset 13 version of Resize where the scale...

Handle ExtractConvBias properly and remove unused bias Quant nodes

Hi @joannapng , thanks for the PR! It would be nice to be able to handle quantized biases with `ExtractConvBias`, but there's a few problems that need to be fixed...

Setting node attributes with wrongly typed values causes bad behaviour

Yes, this can cause all sorts of havoc. Ideally there should be type checking as part of `CustomOp.set_nodeattr`.

[MultiThreshold] Generalize data layouts for node execution

Instead of relying on data layout strings, how about we switch to an attribute `axis` (like many standard ONNX ops do) to indicate the location (index) of the channels axis?...

Add input quantization qkeras converter

Would anyone who is more familiar with the QKeras -> QONNX conversion flow be able to review this? @jmduarte @vloncar @jmitrevs