add

app.py

@@ -62,7 +62,7 @@ def init_leaderboard(dataframe):
     if dataframe is None or dataframe.empty:
         raise ValueError("Leaderboard DataFrame is empty or None.")
 
-    pdb.set_trace()
+    # pdb.set_trace()
     num_cols = dataframe.select_dtypes(include=['float']).columns
     styler = dataframe.style.format({col: "{:.1f}" for col in num_cols})
     return gr.components.Dataframe(

src/leaderboard/read_evals.py

@@ -6,6 +6,7 @@ from dataclasses import dataclass
 
 import dateutil
 import numpy as np
+import pandas as pd
 
 from src.display.formatting import make_clickable_model
 from src.display.utils import AutoEvalColumn, ModelType, Tasks, Precision, WeightType, Metrics
@@ -28,7 +29,7 @@ class EvalResult:
     # architecture: str = "Unknown" 
     # license: str = "?"
     # likes: int = 0
-    num_params: int = None
+    num_params: int = pd.NA
     # date: str = "" # submission date of request file
     # still_on_hub: bool = False
     link: str = ''
@@ -48,7 +49,7 @@ class EvalResult:
         org_and_model = config.get("model_name", config.get("model_args", None))
         org_and_model = org_and_model.split("/", 1)
         link = config.get("link", '')
-        params = config.get("Params", None)
+        params = config.get("Params", pd.NA)
 
         if len(org_and_model) == 1:
             org = None

src/populate.py

@@ -10,6 +10,7 @@ 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, task) -> pd.DataFrame:
     """Creates a dataframe from all the individual experiment results"""
+    # import pdb; pdb.set_trace()
     raw_data = get_raw_eval_results(results_path, requests_path, task)
     all_data_json = [v.to_dict() for v in raw_data]