[Quantization] Calibration for local scales

Hi,

I was wondering if there is any documentation/tutorial regarding the use of local scales while quantizing models?

@vinx13 you mentioned you have a script that can do that? Is this already updated for the current TVM master?

Thanks

The script is quantization_eval.py · GitHub , which is a bit out of date

There is a WIP pr integrating this script into tvm

1 Like