Add support for multiple loggers (#903)

* Add support for multiple loggers

* Fix PEP

* Cleanup

* Cleanup

* Add typing to loggers

* Update base.py

* Replace duck typing with isinstance check

* Update CHANGELOG.md

* Update comet experiment type, Switch to abstractmethod in logging.py

* Fix test

* Add passes to LightningLoggerBase

* Update experiment_logging.rst

CHANGELOG.md

@@ -22,6 +22,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added Tensor Processing Unit (TPU) support ([#868](https://github.com/PyTorchLightning/pytorch-lightning/pull/868))
 - Added semantic segmentation example ([#751](https://github.com/PyTorchLightning/pytorch-lightning/pull/751),[#876](https://github.com/PyTorchLightning/pytorch-lightning/pull/876))
 - Split callbacks in multiple files ([#849](https://github.com/PyTorchLightning/pytorch-lightning/pull/849))
+- Added support for multiple loggers to be passed to `Trainer` as an iterable (e.g. list, tuple, etc.) ([#903](https://github.com/PyTorchLightning/pytorch-lightning/pull/903))
 
 ### Changed
 

docs/source/experiment_logging.rst

@@ -137,3 +137,26 @@ The Wandb logger is available anywhere in your LightningModule
          some_img = fake_image()
          self.logger.experiment.add_image('generated_images', some_img, 0)
 
+
+Multiple Loggers
+^^^^^^^^^^^^^^^^^
+
+PyTorch-Lightning supports use of multiple loggers, just pass a list to the `Trainer`.
+
+.. code-block:: python
+
+   from pytorch_lightning.loggers import TensorBoardLogger, TestTubeLogger
+   
+   logger1 = TensorBoardLogger("tb_logs", name="my_model")
+   logger2 = TestTubeLogger("tt_logs", name="my_model")
+   trainer = Trainer(logger=[logger1, logger2])
+   
+The loggers are available as a list anywhere in your LightningModule
+
+.. code-block:: python
+
+   class MyModule(pl.LightningModule):
+
+      def __init__(self, ...):
+         some_img = fake_image()
+         self.logger.experiment[0].add_image('generated_images', some_img, 0)

pytorch_lightning/loggers/__init__.py

@@ -74,10 +74,10 @@ def any_lightning_module_function_or_hook(...):
 """
 from os import environ
 
-from .base import LightningLoggerBase, rank_zero_only
+from .base import LightningLoggerBase, LoggerCollection, rank_zero_only
 from .tensorboard import TensorBoardLogger
 
-__all__ = ['TensorBoardLogger']
+__all__ = ['TensorBoardLogger', 'LoggerCollection']
 
 try:
     # needed to prevent ImportError and duplicated logs.

pytorch_lightning/loggers/base.py

@@ -1,11 +1,14 @@
-from abc import ABC
+import argparse
+from abc import ABC, abstractmethod
 from functools import wraps
+from typing import Union, Optional, Dict, Iterable, Any, Callable, List
 
 
-def rank_zero_only(fn):
+def rank_zero_only(fn: Callable):
     """Decorate a logger method to run it only on the process with rank 0.
 
-    :param fn: Function to decorate
+    Args:
+        fn: Function to decorate
     """
 
     @wraps(fn)
@@ -23,52 +26,113 @@ def __init__(self):
         self._rank = 0
 
     @property
-    def experiment(self):
-        raise NotImplementedError()
+    @abstractmethod
+    def experiment(self) -> Any:
+        """Return the experiment object associated with this logger"""
+        pass
 
-    def log_metrics(self, metrics, step):
+    @abstractmethod
+    def log_metrics(self, metrics: Dict[str, float], step: Optional[int] = None):
         """Record metrics.
 
-        :param float metric: Dictionary with metric names as keys and measured quanties as values
-        :param int|None step: Step number at which the metrics should be recorded
+        Args:
+            metrics: Dictionary with metric names as keys and measured quantities as values
+            step: Step number at which the metrics should be recorded
         """
-        raise NotImplementedError()
+        pass
 
-    def log_hyperparams(self, params):
+    @abstractmethod
+    def log_hyperparams(self, params: argparse.Namespace):
         """Record hyperparameters.
 
-        :param params: argparse.Namespace containing the hyperparameters
+        Args:
+            params: argparse.Namespace containing the hyperparameters
         """
-        raise NotImplementedError()
+        pass
 
     def save(self):
         """Save log data."""
+        pass
 
-    def finalize(self, status):
+    def finalize(self, status: str):
         """Do any processing that is necessary to finalize an experiment.
 
-        :param status: Status that the experiment finished with (e.g. success, failed, aborted)
+        Args:
+            status: Status that the experiment finished with (e.g. success, failed, aborted)
         """
+        pass
 
     def close(self):
         """Do any cleanup that is necessary to close an experiment."""
+        pass
 
     @property
-    def rank(self):
+    def rank(self) -> int:
         """Process rank. In general, metrics should only be logged by the process with rank 0."""
         return self._rank
 
     @rank.setter
-    def rank(self, value):
+    def rank(self, value: int):
         """Set the process rank."""
         self._rank = value
 
     @property
-    def name(self):
+    @abstractmethod
+    def name(self) -> str:
         """Return the experiment name."""
-        raise NotImplementedError("Sub-classes must provide a name property")
+        pass
 
     @property
-    def version(self):
+    @abstractmethod
+    def version(self) -> Union[int, str]:
         """Return the experiment version."""
-        raise NotImplementedError("Sub-classes must provide a version property")
+        pass
+
+
+class LoggerCollection(LightningLoggerBase):
+    """The `LoggerCollection` class is used to iterate all logging actions over the given `logger_iterable`.
+
+    Args:
+        logger_iterable: An iterable collection of loggers
+    """
+
+    def __init__(self, logger_iterable: Iterable[LightningLoggerBase]):
+        super().__init__()
+        self._logger_iterable = logger_iterable
+
+    @property
+    def experiment(self) -> List[Any]:
+        return [logger.experiment() for logger in self._logger_iterable]
+
+    def log_metrics(self, metrics: Dict[str, float], step: Optional[int] = None):
+        [logger.log_metrics(metrics, step) for logger in self._logger_iterable]
+
+    def log_hyperparams(self, params: argparse.Namespace):
+        [logger.log_hyperparams(params) for logger in self._logger_iterable]
+
+    def save(self):
+        [logger.save() for logger in self._logger_iterable]
+
+    def finalize(self, status: str):
+        [logger.finalize(status) for logger in self._logger_iterable]
+
+    def close(self):
+        [logger.close() for logger in self._logger_iterable]
+
+    @property
+    def rank(self) -> int:
+        return self._rank
+
+    @rank.setter
+    def rank(self, value: int):
+        self._rank = value
+        for logger in self._logger_iterable:
+            logger.rank = value
+
+    @property
+    def name(self) -> str:
+        return '_'.join([str(logger.name) for logger in self._logger_iterable])
+
+    @property
+    def version(self) -> str:
+        return '_'.join([str(logger.version) for logger in self._logger_iterable])

pytorch_lightning/loggers/comet.py

@@ -5,13 +5,15 @@
 CometLogger
 -------------
 """
-
+import argparse
 from logging import getLogger
+from typing import Optional, Union, Dict
 
 try:
     from comet_ml import Experiment as CometExperiment
     from comet_ml import ExistingExperiment as CometExistingExperiment
     from comet_ml import OfflineExperiment as CometOfflineExperiment
+    from comet_ml import BaseExperiment as CometBaseExperiment
     try:
         from comet_ml.api import API
     except ImportError:
@@ -33,9 +35,10 @@ class CometLogger(LightningLoggerBase):
     Log using `comet.ml <https://www.comet.ml>`_.
     """
 
-    def __init__(self, api_key=None, save_dir=None, workspace=None,
-                 rest_api_key=None, project_name=None, experiment_name=None,
-                 experiment_key=None, **kwargs):
+    def __init__(self, api_key: Optional[str] = None, save_dir: Optional[str] = None,
+                 workspace: Optional[str] = None, project_name: Optional[str] = None,
+                 rest_api_key: Optional[str] = None, experiment_name: Optional[str] = None,
+                 experiment_key: Optional[str] = None, **kwargs):
         r"""
 
         Requires either an API Key (online mode) or a local directory path (offline mode)
@@ -77,8 +80,8 @@ def __init__(self, api_key=None, save_dir=None, workspace=None,
             If project name does not already exists Comet.ml will create a new project.
             rest_api_key (str): Optional. Rest API key found in Comet.ml settings.
                 This is used to determine version number
-            experiment_name (str): Optional. String representing the name for this particular experiment on Comet.ml
-
+            experiment_name (str): Optional. String representing the name for this particular experiment on Comet.ml.
+            experiment_key (str): Optional. If set, restores from existing experiment.
         """
         super().__init__()
         self._experiment = None
@@ -120,7 +123,7 @@ def __init__(self, api_key=None, save_dir=None, workspace=None,
                 logger.exception("Failed to set experiment name for comet.ml logger")
 
     @property
-    def experiment(self):
+    def experiment(self) -> CometBaseExperiment:
         r"""
 
         Actual comet object. To use comet features do the following.
@@ -161,11 +164,11 @@ def experiment(self):
         return self._experiment
 
     @rank_zero_only
-    def log_hyperparams(self, params):
+    def log_hyperparams(self, params: argparse.Namespace):
         self.experiment.log_parameters(vars(params))
 
     @rank_zero_only
-    def log_metrics(self, metrics, step=None):
+    def log_metrics(self, metrics: Dict[str, float], step: Optional[int] = None):
         # Comet.ml expects metrics to be a dictionary of detached tensors on CPU
         for key, val in metrics.items():
             if is_tensor(val):
@@ -177,7 +180,7 @@ def reset_experiment(self):
         self._experiment = None
 
     @rank_zero_only
-    def finalize(self, status):
+    def finalize(self, status: str):
         r"""
         When calling self.experiment.end(), that experiment won't log any more data to Comet. That's why, if you need
         to log any more data you need to create an ExistingCometExperiment. For example, to log data when testing your
@@ -190,13 +193,13 @@ def finalize(self, status):
         self.reset_experiment()
 
     @property
-    def name(self):
+    def name(self) -> str:
         return self.experiment.project_name
 
     @name.setter
-    def name(self, value):
+    def name(self, value: str):
         self.experiment.set_name(value)
 
     @property
-    def version(self):
+    def version(self) -> str:
         return self.experiment.id

pytorch_lightning/loggers/mlflow.py

@@ -23,9 +23,10 @@ def any_lightning_module_function_or_hook(...):
         self.logger.experiment.whatever_ml_flow_supports(...)
 
 """
-
+import argparse
 from logging import getLogger
 from time import time
+from typing import Optional, Dict, Any
 
 try:
     import mlflow
@@ -38,7 +39,8 @@ def any_lightning_module_function_or_hook(...):
 
 
 class MLFlowLogger(LightningLoggerBase):
-    def __init__(self, experiment_name, tracking_uri=None, tags=None):
+    def __init__(self, experiment_name: str, tracking_uri: Optional[str] = None,
+                 tags: Dict[str, Any] = None):
         r"""
 
         Logs using MLFlow
@@ -55,7 +57,7 @@ def __init__(self, experiment_name, tracking_uri=None, tags=None):
         self.tags = tags
 
     @property
-    def experiment(self):
+    def experiment(self) -> mlflow.tracking.MlflowClient:
         r"""
 
         Actual mlflow object. To use mlflow features do the following.
@@ -85,12 +87,12 @@ def run_id(self):
         return self._run_id
 
     @rank_zero_only
-    def log_hyperparams(self, params):
+    def log_hyperparams(self, params: argparse.Namespace):
         for k, v in vars(params).items():
             self.experiment.log_param(self.run_id, k, v)
 
     @rank_zero_only
-    def log_metrics(self, metrics, step=None):
+    def log_metrics(self, metrics: Dict[str, float], step: Optional[int] = None):
         timestamp_ms = int(time() * 1000)
         for k, v in metrics.items():
             if isinstance(v, str):
@@ -104,15 +106,15 @@ def save(self):
         pass
 
     @rank_zero_only
-    def finalize(self, status="FINISHED"):
+    def finalize(self, status: str = "FINISHED"):
         if status == 'success':
             status = 'FINISHED'
         self.experiment.set_terminated(self.run_id, status)
 
     @property
-    def name(self):
+    def name(self) -> str:
         return self.experiment_name
 
     @property
-    def version(self):
+    def version(self) -> str:
         return self._run_id