Hi, I ran ViT example and I found that, there is training step after calibration.
I think the ViT exmaple is for PTQ, so I expected it to run only calibration and quantization steps.
But that example actually run training after calibration, so I think it's QAT.
I'm trying to apply adaround to ViT by changing imagenet_example/PTQ/ptq.py with timm's ViT.
But I still get stuck with some errors.
Hi, I ran ViT example and I found that, there is training step after calibration. I think the ViT exmaple is for PTQ, so I expected it to run only calibration and quantization steps. But that example actually run training after calibration, so I think it's QAT.
I'm trying to apply adaround to ViT by changing imagenet_example/PTQ/ptq.py with timm's ViT. But I still get stuck with some errors.
How to do the real PTQ for ViT?
Thank you.