When looking at your config, I see you're using a custom quant_predicate. Would you mind sharing it and how you tested what layers should best be quantized at what settings?
Thanks in advance!
· Sign up or log in to comment
