Do not reset Loops total counters

CHANGELOG.md

@@ -36,6 +36,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
   * Added dataclasses for progress tracking ([#6603](https://github.com/PyTorchLightning/pytorch-lightning/pull/6603), [#7574](https://github.com/PyTorchLightning/pytorch-lightning/pull/7574), [#8140](https://github.com/PyTorchLightning/pytorch-lightning/pull/8140), [#8362](https://github.com/PyTorchLightning/pytorch-lightning/pull/8362))
   * Add `{,load_}state_dict` to the progress tracking dataclasses ([#8140](https://github.com/PyTorchLightning/pytorch-lightning/pull/8140))
   * Connect the progress tracking dataclasses to the loops ([#8244](https://github.com/PyTorchLightning/pytorch-lightning/pull/8244), [#8362](https://github.com/PyTorchLightning/pytorch-lightning/pull/8362))
+  * Do not reset the progress tracking dataclasses total counters ([#8475](https://github.com/PyTorchLightning/pytorch-lightning/pull/8475))
 
 
 - Added support for passing a `LightningDataModule` positionally as the second argument to `trainer.{validate,test,predict}` ([#7431](https://github.com/PyTorchLightning/pytorch-lightning/pull/7431))

pytorch_lightning/loops/base.py

@@ -18,7 +18,7 @@
 from deprecate import void
 
 import pytorch_lightning as pl
-from pytorch_lightning.trainer.progress import BaseProgress, Tracker
+from pytorch_lightning.trainer.progress import BaseProgress, Progress
 from pytorch_lightning.utilities.apply_func import apply_to_collection
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 
@@ -195,11 +195,6 @@ def _load_from_state_dict(self, state_dict: Dict, prefix: str, restart_progress:
             if isinstance(v, BaseProgress):
                 v.load_state_dict(state_dict[prefix + k])
                 if restart_progress:
-
-                    def restart(tracker: Tracker):
-                        tracker.reset_on_restart()
-
-                    apply_to_collection(v, Tracker, restart)
-
+                    apply_to_collection(v, Progress, lambda p: p.current.reset_on_restart())
         self.on_load_checkpoint(state_dict[prefix + "state_dict"])
         self.restarting = True

pytorch_lightning/trainer/progress.py

@@ -65,12 +65,12 @@ def __setattr__(self, key: str, value: int) -> None:
             raise AttributeError(f"The '{key}' attribute is meant to be unused")
         return super().__setattr__(key, value)
 
-    def __repr__(self):
+    def __repr__(self) -> str:
         # hide `None` fields
         args = [f"{k}={v}" for k, v in self.__dict__.items() if v is not None]
         return f"{self.__class__.__name__}({', '.join(args)})"
 
-    def reset_on_restart(self):
+    def reset_on_restart(self) -> None:
         """Reset the progress on restart"""
         value = self.completed if self.processed is None else self.processed
 

tests/loops/test_loops.py

@@ -351,8 +351,8 @@ def val_dataloader(self):
     trainer.fit_loop.load_state_dict(checkpoint)
     expected = {
         "total": {
-            "ready": total_val_batch,
-            "started": total_val_batch,
+            "ready": total_val_batch + 1,
+            "started": total_val_batch + 1,
             "processed": total_val_batch,
             "completed": total_val_batch
         },
@@ -555,6 +555,5 @@ def configure_optimizers_multiple(self):
     trainer.fit_loop.load_state_dict(checkpoint["loops"]["fit_loop"])
     state_dict = trainer.fit_loop.state_dict()
     assert state_dict != checkpoint["loops"]["fit_loop"]
-    # TODO(@carmocca): do not reset for total
-    assert state_dict["epoch_progress"]["total"]["started"] == stop_epoch
+    assert state_dict["epoch_progress"]["total"]["started"] == stop_epoch + 1
     assert state_dict["epoch_progress"]["current"]["started"] == stop_epoch