Perplexity For Language Modeling

While modeling language models, a crucial part is to test its efficiency. Initially I relied on calculating the loss and tried to minimize it. Yes, we need the loss to be minimum to boost the prediction. But is loss the only metric to go for? Will it...