PR: Fix Duplicate Metric Logging in MLFlowLogger to Prevent MLflow Database Errors

requirements/pytorch/test.txt

@@ -17,3 +17,4 @@ fastapi  # for `ServableModuleValidator`  # not setting version as re-defined in
 uvicorn  # for `ServableModuleValidator`  # not setting version as re-defined in App
 
 tensorboard >=2.9.1, <2.20.0  # for `TensorBoardLogger`
+mlflow

src/lightning/pytorch/loggers/mlflow.py

@@ -109,6 +109,12 @@ def any_lightning_module_function_or_hook(self):
         ModuleNotFoundError:
             If required MLFlow package is not installed on the device.
 
+    Note:
+        As of vX.XX, MLFlowLogger will skip logging any metric (same name and step)
+        more than once per run, to prevent database unique constraint violations on
+        some MLflow backends (such as PostgreSQL). Only the first value for each (metric, step)
+        pair will be logged per run. This improves robustness for all users.
+
     """
 
     LOGGER_JOIN_CHAR = "-"
@@ -151,6 +157,7 @@ def __init__(
         from mlflow.tracking import MlflowClient
 
         self._mlflow_client = MlflowClient(tracking_uri)
+        self._logged_metrics = set()  # Track (key, step)
 
     @property
     @rank_zero_experiment
@@ -201,6 +208,7 @@ def experiment(self) -> "MlflowClient":
             resolve_tags = _get_resolve_tags()
             run = self._mlflow_client.create_run(experiment_id=self._experiment_id, tags=resolve_tags(self.tags))
             self._run_id = run.info.run_id
+            self._logged_metrics.clear()
         self._initialized = True
         return self._mlflow_client
 
@@ -266,6 +274,12 @@ def log_metrics(self, metrics: Mapping[str, float], step: Optional[int] = None)
                     category=RuntimeWarning,
                 )
                 k = new_k
+
+            metric_id = (k, step or 0)
+            if metric_id in self._logged_metrics:
+                continue
+            self._logged_metrics.add(metric_id)
+
             metrics_list.append(Metric(key=k, value=v, timestamp=timestamp_ms, step=step or 0))
 
         self.experiment.log_batch(run_id=self.run_id, metrics=metrics_list, **self._log_batch_kwargs)

tests/tests_pytorch/loggers/test_mlflow.py

@@ -427,3 +427,50 @@ def test_set_tracking_uri(mlflow_mock):
     mlflow_mock.set_tracking_uri.assert_not_called()
     _ = logger.experiment
     mlflow_mock.set_tracking_uri.assert_called_with("the_tracking_uri")
+
+
+def test_mlflowlogger_metric_deduplication(monkeypatch):
+    import types
+
+    from lightning.pytorch.loggers.mlflow import MLFlowLogger
+
+    # Dummy MLflow client to record log_batch calls
+    logged_metrics = []
+
+    class DummyMlflowClient:
+        def log_batch(self, run_id, metrics, **kwargs):
+            logged_metrics.extend(metrics)
+
+        def set_tracking_uri(self, uri):
+            pass
+
+        def create_run(self, experiment_id, tags):
+            class Run:
+                info = types.SimpleNamespace(run_id="dummy_run_id")
+
+            return Run()
+
+        def get_run(self, run_id):
+            class Run:
+                info = types.SimpleNamespace(experiment_id="dummy_experiment_id")
+
+            return Run()
+
+        def get_experiment_by_name(self, name):
+            return None
+
+        def create_experiment(self, name, artifact_location=None):
+            return "dummy_experiment_id"
+
+    # Patch the MLFlowLogger to use DummyMlflowClient
+    monkeypatch.setattr("mlflow.tracking.MlflowClient", lambda *a, **k: DummyMlflowClient())
+
+    logger = MLFlowLogger(experiment_name="test_exp")
+    logger.log_metrics({"foo": 1.0}, step=5)
+    logger.log_metrics({"foo": 1.0}, step=5)  # duplicate
+
+    # Only the first metric should be logged
+    assert len(logged_metrics) == 1
+    assert logged_metrics[0].key == "foo"
+    assert logged_metrics[0].value == 1.0
+    assert logged_metrics[0].step == 5