add optional parameters to the wandb logging functionality

det-metrics.py

@@ -341,12 +341,15 @@ class DetectionMetric(evaluate.Metric):
         )
         return fig
 
-    def wandb(self, results , wandb_project='detection_metrics'):
+    def wandb(self, results , wandb_runs: list = None, wandb_section: str = None, wandb_project='detection_metrics'):
         """
         Logs metrics to Weights and Biases (wandb) for tracking and visualization.
     
         This function logs the provided metrics to Weights and Biases (wandb), a platform for tracking machine learning experiments.
         Each key in the `results` dictionary represents a separate run and the corresponding value contains the metrics for that run.
+        If a W&B run list is provided, the results of the runs will be added to the passed W&B runs. Otherwise new W&B runs will be created.
+        If a W&B section ist provided, the metrics will be logged in this section drop-down. Otherwise no extra W&B section is created
+        and the metrics are logged directly.
         The function logs in to wandb using an API key obtained from the secret 'WANDB_API_KEY', initializes a run for 
         each key in `results` and logs the metrics.
     
@@ -357,6 +360,9 @@ class DetectionMetric(evaluate.Metric):
                                 "run1": {"metrics": {"accuracy": 0.9, "loss": 0.1}},
                                 "run2": {"metrics": {"accuracy": 0.85, "loss": 0.15}}
                             }
+            wandb_runs (list, optional): A list containing W&B runs where the results should be added
+                                         (e.g. the first item in results will be added to the first run in wandb_runs, etc.)
+            wandb_section (str, optional): A string to specify the W&B
             wandb_project (str, optional): The name of the wandb project to which the runs will be logged. Defaults to 'detection_metrics'.
     
         Environment Variables:
@@ -375,10 +381,17 @@ class DetectionMetric(evaluate.Metric):
         formatted_datetime = current_datetime.strftime("%Y-%m-%d_%H-%M-%S")
         wandb.login(key=os.getenv('WANDB_API_KEY'))
 
-        for k in results.keys():
-            run = wandb.init(project=wandb_project, name=f"{k}-{formatted_datetime}")
-            run.log(results[k]['metrics'])
-            run.finish()
+        if not wandb_runs is None:
+            assert len(wandb_runs) == len(results), "runs and results must have the same length"
+
+        for i, k in enumerate(results.keys()):
+            if wandb_runs is None:
+                run = wandb.init(project=wandb_project, name=f"{k}-{formatted_datetime}")
+            else:
+                run = wandb_runs[i]
+            run.log({f"{wandb_section}/{m}" : v for m, v in results[k]['metrics'].items()} if wandb_section is not None else results[k]['metrics'])
+            if wandb_runs is None: 
+                run.finish()
 
     def _generate_sample_data(self):
         """