manual accuracy check FAILS

[AnirudhG07]

hello, I need help. I am currently learning basics of pytorch and so I myself tried MNIST numbers data.
So my accuracy came out good. 97% around but when I check accuracy manually by checking calling each data element and comparing actual to predicted number, the accuracy I am getting is <10%

I am sharing codes of checking and some other areas you might need to see.

Pls HELP!

[sametucuncu]

You may use test_loader to iterate through examples for evaluation.
Try to put some print statements between lines to see if you get something wrong. (Especially after scores = model(data) and the one below.)

[AnirudhG07]

Its giving 38% accuracy with data_loader. and decreases every time I run it again and again. while training, the model is evaluating with 97% accuracy.
I gave it an image of mine and it is giving random incorrect answers.
Tell me how to fix that please