Merge branch 'main' of https://huggingface.co/spaces/SEA-AI/user-friendly-metrics into main

user-friendly-metrics.py

@@ -165,6 +165,7 @@ class UserFriendlyMetrics(evaluate.Metric):
         wandb_project="user_friendly_metrics",
         log_plots: bool = True,
         debug: bool = False,
+        log_per_sequence = False
     ):
         """
         Logs metrics to Weights and Biases (wandb) for tracking and visualization, including categorized bar charts for overall metrics.
@@ -190,9 +191,12 @@ class UserFriendlyMetrics(evaluate.Metric):
             self.wandb_run(result = result, 
                     wandb_run_name = wandb_run_name, 
                     wandb_project = wandb_project, 
-                    debug = debug)
+                    debug = debug,
+                    wandb_section = wandb_section,
+                    log_plots = log_plots,
+                    log_per_sequence = log_per_sequence)
 
-    def wandb_run(self, result, wandb_run_name, wandb_project, debug, wandb_section = None, log_plots = True):
+    def wandb_run(self, result, wandb_run_name, wandb_project, debug, wandb_section = None, log_plots = True, log_per_sequence = False):
             
         run = wandb.init(
             project = wandb_project,
@@ -258,24 +262,25 @@ class UserFriendlyMetrics(evaluate.Metric):
                         }
                     )
 
-        if "per_sequence" in result:
-            sorted_sequences = sorted(
-                result["per_sequence"].items(),
-                key=lambda x: next(iter(x[1].values()), {}).get("all", {}).get("f1", 0),
-                reverse=True,  # Set to True for descending order
-            )
-
-            for sequence_name, sequence_data in sorted_sequences:
-                for metric, value in sequence_data["all"].items():
-                    log_key = (
-                        f"{wandb_section}/per_sequence/{sequence_name}/{metric}"
-                        if wandb_section
-                        else f"per_sequence/{sequence_name}/{metric}"
-                    )
-                    run.log({log_key: value})
-                    if debug:
-                        print(f" {log_key} = {value}")
-                print("----------------------------------------------------")
+        if log_per_sequence:
+            if "per_sequence" in result:
+                sorted_sequences = sorted(
+                    result["per_sequence"].items(),
+                    key=lambda x: next(iter(x[1].values()), {}).get("all", {}).get("recall", 0),
+                    reverse=True,  # Set to True for descending order
+                )
+    
+                for sequence_name, sequence_data in sorted_sequences:
+                    for metric, value in sequence_data["all"].items():
+                        log_key = (
+                            f"{wandb_section}/per_sequence/{sequence_name}/{metric}"
+                            if wandb_section
+                            else f"per_sequence/{sequence_name}/{metric}"
+                        )
+                        run.log({log_key: value})
+                        if debug:
+                            print(f" {log_key} = {value}")
+                    print("----------------------------------------------------")
 
         if debug:
             print("\nDebug Mode: Logging Summary and History")