Since it's simulated quantization and the model can't actually be quantized, how do we evaluate the model's performance?I have read many papers, but it seems that I haven't seen any articles explaining how this issue is solved. However, the experimental sections are very detailed.
Since it's simulated quantization and the model can't actually be quantized, how do we evaluate the model's performance?I have read many papers, but it seems that I haven't seen any articles explaining how this issue is solved. However, the experimental sections are very detailed.