[Bug-Fix]:properties `current_epoch` and `global_step` between model and trainer same always (#3785)

* make current_epoch and global_step to be same as trainer, after model restore.

* remove assignment here

* test

* minor modification

* Update pytorch_lightning/core/lightning.py

type check, better clarity

Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>

* Update pytorch_lightning/core/lightning.py

type check, better clarity

Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>

* comments for current_epoch and global_step properties

* Update tests/models/test_restore.py

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

* update comments according to the changes made

* Update tests/models/test_restore.py

* add current_epoch, global_step to jit ignore list

* Add comments to CHANGELOG

* Update CHANGELOG.md

* Update tests/models/test_restore.py

Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

CHANGELOG.md

@@ -103,6 +103,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Fixed aggregation of metrics ([#3517](https://github.com/PyTorchLightning/pytorch-lightning/pull/3517))
 
+- Fixed `current_epoch` and `global_step` properties mismatch between `Trainer` and `LightningModule` ([#3785](https://github.com/PyTorchLightning/pytorch-lightning/pull/3785))
+
 ## [0.9.0] - YYYY-MM-DD
 
 ### Added

pytorch_lightning/core/lightning.py

@@ -68,6 +68,8 @@ class LightningModule(
         "example_input_array",
         "hparams",
         "on_gpu",
+        "current_epoch",
+        "global_step",
     ] + DeviceDtypeModuleMixin.__jit_unused_properties__
 
     def __init__(self, *args, **kwargs):
@@ -79,12 +81,6 @@ class LightningModule(
 
         self.exp_save_path = None
 
-        #: The current epoch
-        self.current_epoch = 0
-
-        #: Total training batches seen across all epochs
-        self.global_step = 0
-
         self.loaded_optimizer_states_dict = {}
 
         #: Pointer to the trainer object
@@ -121,6 +117,16 @@ class LightningModule(
     def example_input_array(self) -> Any:
         return self._example_input_array
 
+    @property
+    def current_epoch(self) -> int:
+        """The current epoch"""
+        return self.trainer.current_epoch if self.trainer else 0
+
+    @property
+    def global_step(self) -> int:
+        """Total training batches seen across all epochs"""
+        return self.trainer.global_step if self.trainer else 0
+
     @example_input_array.setter
     def example_input_array(self, example: Any) -> None:
         self._example_input_array = example

pytorch_lightning/trainer/training_loop.py

@@ -213,8 +213,7 @@ class TrainLoop:
         except Exception:
             pass
 
-        # update training progress in trainer and model
-        model.current_epoch = epoch
+        # update training progress in trainer
         self.trainer.current_epoch = epoch
 
         # changing gradient according accumulation_scheduler
@@ -520,7 +519,6 @@ class TrainLoop:
         should_check_val = False
         for batch_idx, (batch, is_last_batch) in train_dataloader:
             self.trainer.batch_idx = batch_idx
-            model.global_step = self.trainer.global_step
 
             # ------------------------------------
             # TRAINING_STEP + TRAINING_STEP_END

tests/models/test_restore.py

@@ -10,11 +10,52 @@ import torch
 
 import tests.base.develop_pipelines as tpipes
 import tests.base.develop_utils as tutils
-from pytorch_lightning import Trainer
+from pytorch_lightning import Trainer, Callback
 from pytorch_lightning.callbacks import ModelCheckpoint
 from tests.base import EvalModelTemplate, GenericEvalModelTemplate
 
 
+class ModelTrainerPropertyParity(Callback):
+
+    def _check_properties(self, trainer, pl_module):
+        assert trainer.global_step == pl_module.global_step
+        assert trainer.current_epoch == pl_module.current_epoch
+
+    def on_train_start(self, trainer, pl_module):
+        self._check_properties(trainer, pl_module)
+
+    def on_train_batch_start(self, trainer, pl_module, *args, **kwargs):
+        self._check_properties(trainer, pl_module)
+
+    def on_train_batch_end(self, trainer, pl_module, *args, **kwargs):
+        self._check_properties(trainer, pl_module)
+
+    def on_epoch_end(self, trainer, pl_module):
+        self._check_properties(trainer, pl_module)
+
+    def on_train_end(self, trainer, pl_module):
+        self._check_properties(trainer, pl_module)
+
+
+def test_resume_from_checkpoint(tmpdir):
+    """ Test that properties like `current_epoch` and `global_step`
+    in model and trainer are always the same. """
+    model = EvalModelTemplate()
+    checkpoint_callback = ModelCheckpoint(filepath=tmpdir, monitor="early_stop_on", save_last=True)
+    trainer_args = dict(
+        default_root_dir=tmpdir,
+        max_epochs=2,
+        logger=False,
+        early_stop_callback=False,
+        checkpoint_callback=checkpoint_callback,
+        callbacks=[ModelTrainerPropertyParity()]  # this performs the assertions
+    )
+    trainer = Trainer(**trainer_args)
+    trainer.fit(model)
+    trainer = Trainer(**trainer_args, resume_from_checkpoint=str(tmpdir / "last.ckpt"))
+    trainer.fit(model)
+
+
 @pytest.mark.skipif(torch.cuda.device_count() < 2, reason="test requires multi-GPU machine")
 def test_running_test_pretrained_model_distrib_dp(tmpdir):
     """Verify `test()` on pretrained model."""