Accuracy decreases after saving and loading

[JuanFMontesinos (Juan Montesinos)]

Hmm i see. I have no clue. Are you using same batch size?
You can try to compare if loaded state dict equals the one which provides good acc.

You may have a bug choosing when to save / from where to load.

Have a look at this:

What does model.eval() do for batchnorm layer?

Hi Everyone, When doing predictions using a model trained with batchnorm, we should set the model to evaluation model. I have a question that how does the evaluation model affect barchnorm operation? What does evaluation model really do for batchnorm operations? Does the model ignore batchnorm?

Check batchnorm is ok