[Quantization]Does the first layer of input data need to be quantified?

For model quantization, I have several questions:

1.Does the first layer of input data need to be quantified?

  1. For the latest version of the code on github, is there a complete document describing tvm quantization?