Added state attributes for MLFlow Logger (#2160)

* added state attributes for MLFlow Logger * autopep8 fix * added state attributes in args Co-authored-by: Ishan-Kumar2 <[email protected]>
pytorch · Aug 12, 2021 · 327cec5 · 327cec5
1 parent f2b812d
commit 327cec5
Show file tree

Hide file tree

Showing 2 changed files with 40 additions and 1 deletion.
diff --git a/ignite/contrib/handlers/mlflow_logger.py b/ignite/contrib/handlers/mlflow_logger.py
@@ -180,6 +180,20 @@ def global_step_transform(*args, **kwargs):
                 global_step_transform=global_step_transform
             )
 
+        Another example where the State Attributes ``trainer.state.alpha`` and ``trainer.state.beta``
+        are also logged along with the NLL and Accuracy after each iteration:
+
+        .. code-block:: python
+
+            mlflow_logger.attach_output_handler(
+                trainer,
+                event_name=Events.ITERATION_COMPLETED,
+                tag="training",
+                metrics=["nll", "accuracy"],
+                state_attributes=["alpha", "beta"],
+            )
+
+
     Args:
         tag: common title for all produced plots. For example, 'training'
         metric_names: list of metric names to plot or a string "all" to plot all available
@@ -193,6 +207,7 @@ def global_step_transform(*args, **kwargs):
             Default is None, global_step based on attached engine. If provided,
             uses function output as global_step. To setup global step from another engine, please use
             :meth:`~ignite.contrib.handlers.mlflow_logger.global_step_from_engine`.
+        state_attributes: list of attributes of the ``trainer.state`` to plot.
 
     Note:
 
@@ -203,6 +218,8 @@ def global_step_transform(*args, **kwargs):
             def global_step_transform(engine, event_name):
                 return engine.state.get_event_attrib_value(event_name)
 
+    ..  versionchanged:: 0.5.0
+        accepts an optional list of `state_attributes`
     """
 
     def __init__(
@@ -211,8 +228,11 @@ def __init__(
         metric_names: Optional[Union[str, List[str]]] = None,
         output_transform: Optional[Callable] = None,
         global_step_transform: Optional[Callable] = None,
+        state_attributes: Optional[List[str]] = None,
     ) -> None:
-        super(OutputHandler, self).__init__(tag, metric_names, output_transform, global_step_transform)
+        super(OutputHandler, self).__init__(
+            tag, metric_names, output_transform, global_step_transform, state_attributes
+        )
 
     def __call__(self, engine: Engine, logger: MLflowLogger, event_name: Union[str, Events]) -> None:
 

diff --git a/tests/ignite/contrib/handlers/test_mlflow_logger.py b/tests/ignite/contrib/handlers/test_mlflow_logger.py
@@ -188,6 +188,25 @@ def test_output_handler_with_global_step_from_engine():
     )
 
 
+def test_output_handler_state_attrs():
+    wrapper = OutputHandler("tag", state_attributes=["alpha", "beta", "gamma"])
+    mock_logger = MagicMock(spec=MLflowLogger)
+    mock_logger.log_metrics = MagicMock()
+
+    mock_engine = MagicMock()
+    mock_engine.state = State()
+    mock_engine.state.iteration = 5
+    mock_engine.state.alpha = 3.899
+    mock_engine.state.beta = torch.tensor(12.21)
+    mock_engine.state.gamma = torch.tensor([21.0, 6.0])
+
+    wrapper(mock_engine, mock_logger, Events.ITERATION_STARTED)
+
+    mock_logger.log_metrics.assert_called_once_with(
+        {"tag alpha": 3.899, "tag beta": torch.tensor(12.21).item(), "tag gamma 0": 21.0, "tag gamma 1": 6.0,}, step=5,
+    )
+
+
 def test_optimizer_params_handler_wrong_setup():
 
     with pytest.raises(TypeError):