## 🚀 Feature Add a `Trainer.predict()`, similar to `Trainer.test()`, that returns the predictions ### Motivation I want to get prediction, without writing `for batch in ...`, `to(cuda)`, `output.append()` ... ### Pitch Making prediction should be part of a `LightningModule` "system". And it should be easy to add this functionality. ### Alternatives Currently the doc suggests to make prediction like normal PyTorch: https://pytorch-lightning.readthedocs.io/en/latest/trainer.html#deployment-prediction ### Additional context Related forum post: https://forums.pytorchlightning.ai/t/save-test-predictions/549/3