AssertionError: MultiThreshold out_scale must be equal to 1.0 for HLS conversion

[lumawu]

Hello there, I have a problem trying to run the end2end_conv toolflow for a custom net.
Using the end2end_conv_w1a1_example as a template, I got to chapter 3 "Partitioning, Conversion to HLS Layers and Folding" when I encountered the following error on line:

model = model.transform(to_hls.InferThresholdingLayer())
AssertionError: MultiThreshold out_scale must be equal to 1.0 for HLS conversion.

Checking the out_scales of MultiThreshold-Nodes, they remain 2.0 right up until the offending line.
Checking the end2end_conv_example, the following line reduces the out_scale from 2.0 to 1.0 for MultiThreshold nodes:

model = model.transform(ConvertBipolarMatMulToXnorPopcount())

This does not happen with my network however. Here are the things I noticed:

• AbsorbTransposeIntoMultithreshold does not absorb the transposes into multithresholds, as seen in the following two pictures:
  Before: https://imgur.com/a/zd5d3Uh
  After: https://imgur.com/a/jTYYG4L
• ConvertBipolarMatmulToXnorPopcount does not change the out_scale of the Multithresholds

Any insights into the issue would be greatly appreciated. If any additional information is required, I will try to provide it ASAP