Clarify requirements for Trainer.fit(ckpt_path="last") (#19066)

Author: awaelchli

docs/source-pytorch/common/evaluation_intermediate.rst

@@ -41,10 +41,10 @@ To run the test set after training completes, use this method.
     # run full training
     trainer.fit(model)
 
-    # (1) load the best checkpoint automatically (lightning tracks this for you)
+    # (1) load the best checkpoint automatically (lightning tracks this for you during .fit())
     trainer.test(ckpt_path="best")
 
-    # (2) load the last available checkpoint
+    # (2) load the last available checkpoint (only works if `ModelCheckpoint(save_last=True)`)
     trainer.test(ckpt_path="last")
 
     # (3) test using a specific checkpoint

src/lightning/pytorch/trainer/connectors/checkpoint_connector.py

@@ -185,7 +185,7 @@ def _parse_ckpt_path(
                 # not an error so it can be set and forget before the first `fit` run
                 rank_zero_warn(
                     f'.{fn}(ckpt_path="last") is set, but there is no last checkpoint available.'
-                    " No checkpoint will be loaded."
+                    " No checkpoint will be loaded. HINT: Set `ModelCheckpoint(..., save_last=True)`."
                 )
                 return None
             ckpt_path = max(candidates_ts, key=candidates_ts.get)  # type: ignore[arg-type]