Batch size in actual testing


I created a CNN, trained it on my dataset and saved the model with the best validation accuracy.
Now I want to actually use it, i.e. I want it to classify images for me.
I am wondering which batch size I should use, because I read somewhere that I should use a batch size of 1. I am using 1 GPU for the training

Any help is appreciated!


Now that you want to use your model (training is over), you should use it in eval mode (call model.eval() before processing your new images). In that case the batch size has no impact so you should set it as large as possible (depending on the memory available on your GPU) to make the process as fast as possible.

Hope this help.

That helped alot, because I totally forgot about my batchnorm.

Thank you very much!!