remove average_pb

src/display/utils.py

@@ -27,7 +27,7 @@ auto_eval_column_dict = []
 auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
 #Scores
 auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Avg. Correlation ⬆️", "number", True)])
-auto_eval_column_dict.append(["average_pb", ColumnContent, ColumnContent("Positional Bias Impact", "number", True)])
+# auto_eval_column_dict.append(["average_pb", ColumnContent, ColumnContent("Positional Bias Impact", "number", True)])
 for task in Tasks:
     auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
 # Model information

src/leaderboard/read_evals.py

@@ -114,11 +114,7 @@ class EvalResult:
     def to_dict(self, mina=0, maxa=1):
         """Converts the Eval Result to a dict compatible with our dataframe display"""
         average = sum([self.results["apcc"], self.results["mpcc"]]) / 2
-        print("self.results mpcc_delta", self.results["mpcc_delta"])
     
-        norm_mpcc_delta = (float(self.results["mpcc_delta"]) - mina) / (maxa - mina)
-        print("norm_mpcc_delta", norm_mpcc_delta)
-        average_pb = sum([norm_mpcc_delta, self.results["mpcc_cons"], self.results["pcon_ab"]]) / 3
         data_dict = {
             "eval_name": self.eval_name,  # not a column, just a save name,
             AutoEvalColumn.precision.name: self.precision.value.name,
@@ -129,7 +125,6 @@ class EvalResult:
             AutoEvalColumn.model.name: make_clickable_model(self.full_model),
             AutoEvalColumn.revision.name: self.revision,
             AutoEvalColumn.average.name: average,
-            AutoEvalColumn.average_pb.name: average_pb,
             AutoEvalColumn.license.name: self.license,
             AutoEvalColumn.likes.name: self.likes,
             AutoEvalColumn.params.name: self.num_params,
@@ -201,11 +196,9 @@ def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResu
             eval_results[eval_name] = eval_result
 
     results = []
-    mina = min([a.results["mpcc_delta"] for a in eval_results.values()])
-    maxa = max([a.results["mpcc_delta"] for a in eval_results.values()])
     for v in eval_results.values():
         try:
-            v.to_dict(mina, maxa)  # we test if the dict version is complete
+            v.to_dict()  # we test if the dict version is complete
             results.append(v)
         except KeyError as e:  # not all eval values present
             print("e", e)

src/populate.py

@@ -11,9 +11,7 @@ from src.leaderboard.read_evals import get_raw_eval_results
 def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
     """Creates a dataframe from all the individual experiment results"""
     raw_data = get_raw_eval_results(results_path, requests_path)
-    mina = min([a.results["mpcc_delta"] for a in raw_data.values()])
-    maxa = max([a.results["mpcc_delta"] for a in raw_data.values()])
-    all_data_json = [v.to_dict(mina, maxa) for v in raw_data]
+    all_data_json = [v.to_dict() for v in raw_data]
 
     df = pd.DataFrame.from_records(all_data_json)
     df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
@@ -30,10 +28,10 @@ def set_style_for_leaderboard_df(df: pd.DataFrame) -> pd.DataFrame:
     # Adding CSS to style the specific column header
     styled_df.set_table_styles({
         AutoEvalColumn.average.name: [{'selector': 'th.col_heading.level0', 'props': 'color: green;'}],
-        AutoEvalColumn.average_pb.name: [{'selector': 'th.col_heading.level0', 'props': 'color: green;'}]
+        # AutoEvalColumn.average_pb.name: [{'selector': 'th.col_heading.level0', 'props': 'color: green;'}]
     }, overwrite=False)
 
-    styled_df.format(na_rep="").bar(align=0,  subset=[AutoEvalColumn.average.name, AutoEvalColumn.average_pb.name], cmap="PiYG")
+    styled_df.format(na_rep="").bar(align=0,  subset=[AutoEvalColumn.average.name], cmap="PiYG")
     return styled_df
 
 def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]: