To fix #3277

ignite/handlers/mlflow_logger.py

@@ -1,5 +1,6 @@
 """MLflow logger and its helper handlers."""
 
+import re
 import warnings
 from typing import Any, Callable, List, Optional, Union
 
@@ -245,15 +246,15 @@ def __call__(self, engine: Engine, logger: MLflowLogger, event_name: Union[str,
             )
 
         # Additionally recheck metric names as MLflow rejects non-valid names with MLflowException
-        from mlflow.utils.validation import _VALID_PARAM_AND_METRIC_NAMES
+        from mlflow.utils.validation import validate_param_and_metric_name
 
         metrics = {}
         for keys, value in rendered_metrics.items():
             key = " ".join(keys)
             metrics[key] = value
 
         for key in list(metrics.keys()):
-            if not _VALID_PARAM_AND_METRIC_NAMES.match(key):
+            if not validate_param_and_metric_name(key):
                 warnings.warn(
                     f"MLflowLogger output_handler encountered an invalid metric name '{key}' that "
                     "will be ignored and not logged to MLflow"

requirements-dev.txt

@@ -17,7 +17,7 @@ tensorboardX
 visdom
 polyaxon
 wandb
-mlflow<2.16.0  # https://github.com/pytorch/ignite/issues/3277
+mlflow
 neptune-client>=0.16.17
 tensorboard
 torchvision