brevitas icon indicating copy to clipboard operation
brevitas copied to clipboard

Published 20 hours ago • verified publisher icon Xilinx

FlexML quantized AvgPool update

Open volcacius opened this issue 1 year ago • 1 comments

Switch over to a Vitis-AI like implementation and export flow for QuantAvgPool2d for FlexML graph quantization. Keep the conversion of AvgPool to dephwise Conv as an option but simplify the flow by having a single scalar weight in the conv layer and by directly doing a AvgPool->QuantConv without passing through float Conv.

@cmcgirr-amd

volcacius avatar Aug 05 '22 15:08 volcacius

Just an note (don't know if you're aware), I tried using a newer torch version than 1.10.0 and the export fails due to the onnx export API having changed keyword arguments

cmcgirr-amd avatar Aug 23 '22 08:08 cmcgirr-amd