update logging docs and decorators (#4431)

* update logging docs

* experiment

* add decorators to base and csv logger methods

* fix

* doc fix

* update docs

* update docs

* Update pytorch_lightning/loggers/base.py

Co-authored-by: chaton <[email protected]>

Author: rohitgr7

docs/source/logging.rst

@@ -124,24 +124,28 @@ Once your training starts, you can view the logs by using your favorite logger o
 Make a custom logger
 ********************
 
-You can implement your own logger by writing a class that inherits from
-:class:`LightningLoggerBase`. Use the :func:`~pytorch_lightning.loggers.base.rank_zero_only`
-decorator to make sure that only the first process in DDP training logs data.
+You can implement your own logger by writing a class that inherits from :class:`~pytorch_lightning.loggers.base.LightningLoggerBase`.
+Use the :func:`~pytorch_lightning.loggers.base.rank_zero_experiment` and :func:`~pytorch_lightning.utilities.distributed.rank_zero_only` decorators to make sure that only the first process in DDP training creates the experiment and logs the data respectively.
 
 .. testcode::
 
     from pytorch_lightning.utilities import rank_zero_only
     from pytorch_lightning.loggers import LightningLoggerBase
+    from pytorch_lightning.loggers.base import rank_zero_experiment
 
     class MyLogger(LightningLoggerBase):
 
+        @property
         def name(self):
             return 'MyLogger'
 
+        @property
+        @rank_zero_experiment
         def experiment(self):
             # Return the experiment object associated with this logger.
             pass
-          
+         
+        @property 
         def version(self):
             # Return the experiment version, int or str.
             return '0.1'
@@ -158,6 +162,7 @@ decorator to make sure that only the first process in DDP training logs data.
             # your code to record metrics goes here
             pass
 
+        @rank_zero_only
         def save(self):
             # Optional. Any code necessary to save logger data goes here
             # If you implement this, remember to call `super().save()`

pytorch_lightning/loggers/base.py

@@ -29,6 +29,17 @@
 from pytorch_lightning.utilities import rank_zero_only
 
 
+def rank_zero_experiment(fn: Callable) -> Callable:
+    """ Returns the real experiment on rank 0 and otherwise the DummyExperiment. """
+    @wraps(fn)
+    def experiment(self):
+        @rank_zero_only
+        def get_experiment():
+            return fn(self)
+        return get_experiment() or DummyExperiment()
+    return experiment
+
+
 class LightningLoggerBase(ABC):
     """
     Base class for experiment loggers.
@@ -410,9 +421,11 @@ def __init__(self):
     def experiment(self):
         return self._experiment
 
+    @rank_zero_only
     def log_metrics(self, metrics, step):
         pass
 
+    @rank_zero_only
     def log_hyperparams(self, params):
         pass
 
@@ -477,14 +490,3 @@ def merge_dicts(
             d_out[k] = (fn or default_func)(values_to_agg)
 
     return d_out
-
-
-def rank_zero_experiment(fn: Callable) -> Callable:
-    """ Returns the real experiment on rank 0 and otherwise the DummyExperiment. """
-    @wraps(fn)
-    def experiment(self):
-        @rank_zero_only
-        def get_experiment():
-            return fn(self)
-        return get_experiment() or DummyExperiment()
-    return experiment

pytorch_lightning/loggers/csv_logs.py

@@ -29,7 +29,7 @@
 
 from pytorch_lightning import _logger as log
 from pytorch_lightning.core.saving import save_hparams_to_yaml
-from pytorch_lightning.loggers.base import LightningLoggerBase
+from pytorch_lightning.loggers.base import LightningLoggerBase, rank_zero_experiment
 from pytorch_lightning.utilities.distributed import rank_zero_only, rank_zero_warn
 
 
@@ -162,6 +162,7 @@ def save_dir(self) -> Optional[str]:
         return self._save_dir
 
     @property
+    @rank_zero_experiment
     def experiment(self) -> ExperimentWriter:
         r"""
 

pytorch_lightning/loggers/wandb.py

@@ -170,6 +170,7 @@ def version(self) -> Optional[str]:
         # don't create an experiment if we don't have one
         return self._experiment.id if self._experiment else self._id
 
+    @rank_zero_only
     def finalize(self, status: str) -> None:
         # offset future training logged on same W&B run
         if self._experiment is not None: