Understanding the difference between on_validation_end and on_validation_epoch_end

[Northo]

What is the intended difference between on_validation_end and on_validation_epoch_end?

As far as I've been able to understand

• on_validation_end does not allow logging, on_validation_epoch_end does
• on_validation_end is where early stopping is being handled
• they both have the same signature
• they are always called in succession (on_validation_epoch_end -> on_validation_end).

Any input on this is greatly appreciated. Mostly

1. Is there an "official" intended difference between the two?
2. What is the best source for understanding when and how hooks are called?
   From the docs, it is clear what hooks exists, but I find it difficult to find clear documentation on when and "how" they are called.

Example code

import lightning as L
from lightning.pytorch.demos.boring_classes import BoringModel, BoringDataModule

class DebugCallback(L.Callback):
    def on_validation_epoch_end(self, trainer, pl_module):
        print("On validation epoch end")
        self.log("foo", 3)

    def on_validation_end(self, trainer, pl_module):
        print("On validation end")
        # self.log("bar", 4)  # <-- raises MisconfigurationException

model = BoringModel()
datamodule = BoringDataModule()

trainer = L.Trainer(
    callbacks=[DebugCallback()],
    max_epochs=2,
)
trainer.fit(model, datamodule=datamodule)
trainer.validate(model, datamodule=datamodule)