05-28-2021 01:05 AM - edited 05-28-2021 01:07 AM
Hi Xilinx Community,
I have been trying to quantize a PyTorch model which uses "split()" layer as one of the layers to split out the features in chunks. Essentially, this layer: https://pytorch.org/docs/stable/generated/torch.split.html.
I am getting the following error during the quantization from the docker container:
"aten_op 'split' parse failed(unsupported)"
Does this mean that the layer is not currently supported by the quantizer?
Could anyone please confirm that this layer is not supported by pytorch quantizer currently? I am using Vitis-AI docker tag 1.3 for this quantization.