[Quantization][AutoTVM] Performance degradation of Quantization + AutoTVM vs FP on x86

Troubleshooting
7

You mean setting this parameter like this: global_scale=8.0,? Or what do you mean by scales?