Fix null tracker metadata and disable mlflow by default (#18)

* fix: return empty tracker metadata when no tracker

* chore: mlflow disabled by default

* chore: update changelog

* Merge develop into fix/null-tracker

CHANGELOG.md

@@ -59,6 +59,7 @@ Keep it human-readable, your future self will thank you!
  - Adjusted PyPI release infrastructure to common ECMWF workflow
  - Bumped versions in Pre-commit hooks
  - Fix crash when logging hyperparameters with missing values in the config
+ - Fixed "null" tracker metadata when tracking is disabled, now returns an empty dict
 
 ### Removed
  - Dependency on mlflow-export-import

src/anemoi/training/config/diagnostics/eval_rollout.yaml

@@ -78,7 +78,7 @@ log:
   tensorboard:
     enabled: False
   mlflow:
-    enabled: True
+    enabled: False
     offline: False
     authentication: False
     log_model: False

src/anemoi/training/diagnostics/callbacks/__init__.py

@@ -789,10 +789,9 @@ def tracker_metadata(self, trainer: pl.Trainer) -> dict:
                     "url": run.url,
                     "project": run.project,
                 }
-            else:
-                self._tracker_metadata = {}
+            return {self._tracker_name: self._tracker_metadata}
 
-        elif self.config.diagnostics.log.mlflow.enabled:
+        if self.config.diagnostics.log.mlflow.enabled:
             self._tracker_name = "mlflow"
 
             from anemoi.training.diagnostics.mlflow.logger import AnemoiMLflowLogger
@@ -808,10 +807,9 @@ def tracker_metadata(self, trainer: pl.Trainer) -> dict:
                     "url": run.info.artifact_uri,
                     "project": run.info.experiment_id,
                 }
-            else:
-                self._tracker_metadata = {}
+            return {self._tracker_name: self._tracker_metadata}
 
-        return {self._tracker_name: self._tracker_metadata}
+        return {}
 
     def _save_checkpoint(self, trainer: pl.Trainer, lightning_checkpoint_filepath: str) -> None:
         if trainer.is_global_zero: