Hi all,
I’m really new to TVM. I’m using TVM to quantize self-designed model. According to the tutorial(https://tvm.apache.org/docs/tutorials/frontend/deploy_quantized.html#sphx-glr-tutorials-frontend-deploy-quantized-py), I can use “kl_divergence” or “global_scale” for calibration mode. For “global_scale” mode, how can I set proper value for “global_scale”? I saw the default value is 8.0, but how is this chosen? Is there any docs that explain the meaning of the parameters?
Thanks.