Resolve FitLoop setter TODOs

src/lightning/pytorch/CHANGELOG.md

@@ -326,6 +326,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Removed `ProgressBarBase.{train_batch_idx,val_batch_idx,test_batch_idx,predict_batch_idx}` properties ([#16760](https://github.com/Lightning-AI/lightning/pull/16760))
 
 
+- Removed the `fit_loop.{min,max}_steps` setters ([#16803](https://github.com/Lightning-AI/lightning/pull/16803))
+
 
 - Removed the `Trainer(track_grad_norm=...)` argument ([#16745](https://github.com/Lightning-AI/lightning/pull/16745))
 

src/lightning/pytorch/loops/fit_loop.py

@@ -102,27 +102,11 @@ def min_steps(self) -> Optional[int]:
         """Returns the minimum number of steps to run."""
         return self.epoch_loop.min_steps
 
-    @min_steps.setter
-    def min_steps(self, value: Optional[int]) -> None:
-        """Sets the minimum number of steps (forwards to epoch_loop)"""
-        # TODO: This setter is required by debugging connector (fast dev run), should be avoided
-        self.epoch_loop.min_steps = value
-
     @property
     def max_steps(self) -> int:
         """Returns the maximum number of steps to run."""
         return self.epoch_loop.max_steps
 
-    @max_steps.setter
-    def max_steps(self, value: int) -> None:
-        """Sets the maximum number of steps (forwards to epoch_loop)"""
-        # TODO: This setter is required by debugging connector (fast dev run), should be avoided
-        if value < -1:
-            raise MisconfigurationException(
-                f"`max_steps` must be a non-negative integer or -1 (infinite steps). You passed in {value}."
-            )
-        self.epoch_loop.max_steps = value
-
     @_Loop.restarting.setter
     def restarting(self, restarting: bool) -> None:
         # if the last epoch completely finished, we are not actually restarting

src/lightning/pytorch/trainer/setup.py

@@ -72,7 +72,7 @@ def _init_debugging_flags(
 
         trainer.limit_test_batches = num_batches
         trainer.limit_predict_batches = num_batches
-        trainer.fit_loop.max_steps = num_batches
+        trainer.fit_loop.epoch_loop.max_steps = num_batches
         trainer.num_sanity_val_steps = 0
         trainer.fit_loop.max_epochs = 1
         trainer.val_check_interval = 1.0

src/lightning/pytorch/tuner/batch_size_scaling.py

@@ -129,7 +129,7 @@ def __scale_batch_reset_params(trainer: "pl.Trainer", steps_per_trial: int) -> N
     if isinstance(loop, pl.loops._FitLoop):
         trainer.limit_train_batches = 1.0
         trainer.limit_val_batches = steps_per_trial
-        trainer.fit_loop.max_steps = steps_per_trial
+        trainer.fit_loop.epoch_loop.max_steps = steps_per_trial
     elif isinstance(loop, pl.loops._EvaluationLoop):
         stage = trainer.state.stage
         assert stage is not None
@@ -145,7 +145,7 @@ def __scale_batch_restore_params(trainer: "pl.Trainer", params: Dict[str, Any])
     loop = trainer._active_loop
     assert loop is not None
     if isinstance(loop, pl.loops._FitLoop):
-        loop.max_steps = params["max_steps"]
+        loop.epoch_loop.max_steps = params["max_steps"]
         trainer.limit_train_batches = params["limit_train_batches"]
         trainer.limit_val_batches = params["limit_val_batches"]
     elif isinstance(loop, pl.loops._EvaluationLoop):

src/lightning/pytorch/tuner/lr_finder.py

@@ -320,7 +320,7 @@ def __lr_finder_reset_params(trainer: "pl.Trainer", num_training: int, early_sto
     # No logging
     trainer.logger = DummyLogger() if trainer.logger is not None else None
     # Max step set to number of iterations starting at current number of iterations
-    trainer.fit_loop.max_steps = num_training + trainer.global_step
+    trainer.fit_loop.epoch_loop.max_steps = num_training + trainer.global_step
     trainer.limit_val_batches = num_training
 
 
@@ -329,10 +329,10 @@ def __lr_finder_restore_params(trainer: "pl.Trainer", params: Dict[str, Any]) ->
     trainer.strategy.lr_scheduler_configs = params["lr_scheduler_configs"]
     trainer.callbacks = params["callbacks"]
     trainer.loggers = params["loggers"]
-    trainer.fit_loop.max_steps = params["max_steps"]
+    loop = trainer.fit_loop
+    loop.epoch_loop.max_steps = params["max_steps"]
     trainer.limit_val_batches = params["limit_val_batches"]
 
-    loop = trainer.fit_loop
     loop.load_state_dict(deepcopy(params["loop_state_dict"]))
     loop.restarting = False
     trainer.should_stop = False

tests/tests_pytorch/loops/test_training_loop.py

@@ -152,11 +152,11 @@ def test_fit_loop_done_log_messages(caplog):
     epoch_loop = Mock()
     epoch_loop.global_step = 10
     fit_loop.epoch_loop = epoch_loop
-    fit_loop.max_steps = 10
+    epoch_loop.max_steps = 10
     assert fit_loop.done
     assert "max_steps=10` reached" in caplog.text
     caplog.clear()
-    fit_loop.max_steps = 20
+    epoch_loop.max_steps = 20
 
     fit_loop.epoch_progress.current.processed = 3
     fit_loop.max_epochs = 3

tests/tests_pytorch/trainer/test_dataloaders.py

@@ -1184,7 +1184,7 @@ def test_dataloaders_reset_and_attach(tmpdir):
     assert trainer.train_dataloader.dataset is dataloader_0.dataset
     assert trainer.val_dataloaders[0].dataset is dataloader_1.dataset
     # 2nd fit
-    trainer.fit_loop.max_steps += 1
+    trainer.fit_loop.epoch_loop.max_steps += 1
     trainer.fit(model, train_dataloaders=dataloader_2, val_dataloaders=dataloader_3)
     assert trainer.train_dataloader.dataset is dataloader_2.dataset
     assert trainer.val_dataloaders[0].dataset is dataloader_3.dataset