Batchsize of QAT

[sf-Liu]

Hi, I'm trying the QAT.

When I do the training, I have to make the batch size more than 1, or the BatchNorm will report errors.
But my target is to generate a quantized model with only 1 input batch, is there any way to change the input batch size of sim.model? Or any method to satisfy my target?