The model's inference result is wrong after using the static quantization of PTQ

The model’s inference result is wrong after using the static quantization of PTQ. I don’t know why. Maybe adding more dataset is useful? Currently, 78 groups are applied