Low accuracy after normalizing dataset for Perceptron - Week 2 perceptron learning practical doubt

During practical session on perceptron learning. Why does the instructor not normalized data?
As when I perform perceptron learning without normalizing the data then I am getting 96% accuracy but when I normalized data using MinMaxScalar as said in lectures (but not done during practical) I am getting only 33% accuracy.
I had applied MinMaxScalar directly on initial X then after I splitted into Xtrain and Xtest

Hi @pritul_dave,
This seems somewhat unusual, can you please share your notebook?

https://colab.research.google.com/drive/1M4Q6L33WihhtGVOKUUQTjbPeNFXxlMkd?usp=sharing