How to access integer weights and scales after quantization

Questions
#1

I use

mod = relay.quantize.quantize(mod, params, dataset=calibrate_dataset())

to quantize a network. However, after the network is quantized, mod and params still seem to be float32 after I print them. Is there a way to see quantized weights and scales like in TFLite or PyTorch?

#2

The actual quantization of weights happens during constant evaluation, so you need to run relay.build first. The output params in json, lib, params = relay.build(...) should be quantized.

#3

Thank you. I found the same answer in another post too.