Fixed fillna anonymous, moved data fetching to utils.py

app.py

@@ -1,9 +1,9 @@
+import os
 import pandas as pd
 import gradio as gr
 from gradio.themes.utils import sizes
 from gradio_leaderboard import Leaderboard
 from dotenv import load_dotenv
-import contextlib
 
 load_dotenv()  # Load environment variables from .env file
 
@@ -21,7 +21,7 @@ from constants import (
     SLACK_URL,
 )
 from submit import make_submission
-from utils import fetch_hf_results, show_output_box
+from utils import fetch_hf_results, show_output_box, periodic_data_fetch
 
 
 def format_leaderboard_table(df_results: pd.DataFrame, assay: str | None = None):
@@ -67,38 +67,16 @@ def get_leaderboard_object(assay: str | None = None):
     return lb
 
 
-# Initialize global dataframe
-fetch_hf_results()
-current_dataframe = pd.read_csv("debug-current-results.csv")
-
-
 def refresh_overall_leaderboard():
+    if not os.path.exists("debug-current-results.csv"):
+        fetch_hf_results()  # Hope this doesn't cause race conditions with the main fetch_hf_results() thread
     current_dataframe = pd.read_csv("debug-current-results.csv")
     return format_leaderboard_table(df_results=current_dataframe)
 
 
-def fetch_latest_data(stop_event):
-    import time
-
-    while not stop_event.is_set():
-        try:
-            fetch_hf_results()
-        except Exception as e:
-            print(f"Error fetching latest data: {e}")
-        time.sleep(3)  # Fetch every 60 seconds
-    print("Exiting data fetch thread")
-
-
-@contextlib.asynccontextmanager
-async def periodic_data_fetch(app):
-    import threading
-
-    event = threading.Event()
-    t = threading.Thread(target=fetch_latest_data, args=(event,), daemon=True)
-    t.start()
-    yield
-    event.set()
-    t.join(3)
+# Initialize global dataframe
+fetch_hf_results()
+current_dataframe = pd.read_csv("debug-current-results.csv")
 
 
 # Make font size bigger using gradio theme

utils.py

@@ -1,3 +1,4 @@
+import contextlib
 from datetime import datetime, timezone, timedelta
 import hashlib
 import os
@@ -66,8 +67,9 @@ def fetch_hf_results():
     df.loc[df["user"].isin(BASELINE_USERNAMES), "user"] = "Baseline"
     # Note: Could optionally add a column "is_baseline" to the dataframe to indicate whether the model is a baseline model or not. If things get crowded.
     # Anonymize the user column at this point (so note: users can submit anonymous / non-anonymous and we'll show their latest submission regardless)
-    anon_rows = df["anonymous"].fillna(True)
-    df.loc[anon_rows, "user"] = "anon-" + df.loc[anon_rows, "user"].apply(readable_hash)
+    df.loc[df["anonymous"], "user"] = "anon-" + df.loc[df["anonymous"], "user"].apply(
+        readable_hash
+    )
 
     # Compare to previous dataframe
     if os.path.exists("debug-current-results.csv"):
@@ -78,6 +80,30 @@ def fetch_hf_results():
     df.to_csv("debug-current-results.csv", index=False)
 
 
+def fetch_latest_data(stop_event):
+    import time
+
+    while not stop_event.is_set():
+        try:
+            fetch_hf_results()
+        except Exception as e:
+            print(f"Error fetching latest data: {e}")
+        time.sleep(3)  # Fetch every 60 seconds
+    print("Exiting data fetch thread")
+
+
+@contextlib.asynccontextmanager
+async def periodic_data_fetch(app):
+    import threading
+
+    event = threading.Event()
+    t = threading.Thread(target=fetch_latest_data, args=(event,), daemon=True)
+    t.start()
+    yield
+    event.set()
+    t.join(3)
+
+
 # Readable hashing function similar to coolname or codenamize
 ADJECTIVES = [
     "ancient",