In a multi-label classification problem, how is the evaluation of one sample and then all samples in detail?

Question

Having a multi-label classification problem, how is the evaluation of each sample? Each individual label has its own part or it should be looked as whole? Giving an example