fix: get MLFLOW_TRACKING_UTI from env variables as default (#2107)

* fix: get MLFLOW_TRACKING_UTI from env variables as default

In `AnomalibMLFlowLogger`, `tracking_uri` was set to None, which was overriding the parameter in the underlying pytorch `MLFloeLogger` so never could log into remote server. This fixes it.

* Update src/anomalib/loggers/mlflow.py

* Update mlflow.py

---------

Co-authored-by: Carlos Nácher <[email protected]>
Co-authored-by: Samet Akcay <[email protected]>

Author: 3 people

src/anomalib/loggers/mlflow.py

@@ -1,5 +1,6 @@
 """MLFlow logger with add image interface."""
 
+import os
 from typing import Literal
 
 import numpy as np
@@ -65,7 +66,7 @@ def __init__(
         self,
         experiment_name: str | None = "anomalib_logs",
         run_name: str | None = None,
-        tracking_uri: str | None = None,
+        tracking_uri: str | None = os.getenv("MLFLOW_TRACKING_URI"),
         save_dir: str | None = "./mlruns",
         log_model: Literal[True, False, "all"] | None = False,
         prefix: str | None = "",