Interim LB v1

app.py

@@ -110,7 +110,7 @@ def update_current_dataframe():
         logger.info("Fetching latest dataset for leaderboard...")
         current_df, current_df_raw = fetch_dataset_df()
         logger.debug(f"Dataset version updated")
-        time.sleep(30)  # Check for updates every 30 sec
+        time.sleep(60)  # Check for updates every 60 sec
 
 threading.Thread(target=update_current_dataframe, daemon=True).start()
 
@@ -119,7 +119,7 @@ threading.Thread(target=update_current_dataframe, daemon=True).start()
 
 with gr.Blocks(title="OpenADMET ADMET Challenge", fill_height=False,
                theme=gr.themes.Default(text_size=sizes.text_lg)) as demo:
-    timer = gr.Timer(10)  # Run every 10 seconds
+    timer = gr.Timer(30)  # Run every 30 seconds
     data_version = gr.State(value=0)
     def increment_data_version(current_version):
         logger.debug("Incrementing data version counter... to " + str(current_version + 1))
@@ -137,8 +137,8 @@ with gr.Blocks(title="OpenADMET ADMET Challenge", fill_height=False,
                 Go to the **Leaderboard** to check out how the challenge is going. 
                 To participate, head out to the **Submit** tab and upload your results as a `CSV` file.
                     
-                **We are releasing an intermediate leaderboard on December 1st so stay tuned!
-                        The submission cutoff for this leaderboard will be November 30th at 6pm EST. **
+                We are releasing an intermediate leaderboard on December 2nd so stay tuned!
+                        The submission cutoff for this leaderboard will be **Dec 1st at 11:59PM GMT.**
                         
                 """
                 )
@@ -307,7 +307,7 @@ with gr.Blocks(title="OpenADMET ADMET Challenge", fill_height=False,
                     select_columns=LB_AVG,
                     search_columns=["user"],
                     render=True,
-                    every=30,
+                    every=60,
                 )
                 # Set up button to download leaderboard as csv file 
                 download_lb = gr.DownloadButton(
@@ -330,7 +330,7 @@ with gr.Blocks(title="OpenADMET ADMET Challenge", fill_height=False,
                         select_columns=LB_COLS,
                         search_columns=["user"],
                         render=True,
-                        every=30,
+                        every=60,
                     )
             # Auto-refresh 
             def refresh_if_changed():

final_lb.py

@@ -5,7 +5,7 @@ from utils import (
     map_metric_to_stats,
     fetch_dataset_df,
 )
-from about import ENDPOINTS, LB_COLS, LB_AVG, results_repo_test, results_repo_validation
+from about import ENDPOINTS, LB_COLS, LB_AVG, results_repo_test
 
 from loguru import logger
 import pandas as pd
@@ -14,7 +14,7 @@ from pathlib import Path
 
 ALL_EPS = ['Average'] + ENDPOINTS
 
-def build_leaderboard(df_results, df_results_raw):
+def build_leaderboard(df_results, df_results_raw, avg_only=True):
     per_ep = {}
     for ep in ALL_EPS:
         df = df_results[df_results["Endpoint"] == ep].copy()
@@ -67,6 +67,8 @@ def build_leaderboard(df_results, df_results_raw):
             per_ep[ep] = avg_leaderboard_clean[avg_cols]
 
         else:
+            if avg_only:
+                continue
             # Delete duplicate entries before sorting (fixing case-sensitive duplicate check)
             df['hf_username'] = df['hf_username'].apply(lambda s: s.lower())
             df = df.sort_values(by="submission time", ascending=False, kind="stable")
@@ -97,14 +99,14 @@ def validate_model_details(tag):
     else:
         return safe_tag
 
-def prepare_lb_csv(save_folder:str):
+def prepare_lb_csv(save_folder:str, avg_only:bool):
     logger.info("Fetching data")
     df_latest, df_latest_raw = fetch_dataset_df(
         download_raw=True, 
         test_repo=results_repo_test
     )
     logger.info("Building leaderboard")
-    per_ep_df = build_leaderboard(df_latest, df_latest_raw)
+    per_ep_df = build_leaderboard(df_latest, df_latest_raw, avg_only)
     logger.info("Saving leaderboard")
     for ep in ALL_EPS:
         df_lb = per_ep_df[ep]
@@ -113,4 +115,4 @@ def prepare_lb_csv(save_folder:str):
     return
 
 if __name__ == "__main__":
-    prepare_lb_csv("intermediate_lbs")  
+    prepare_lb_csv("intermediate_lbs", avg_only=True)  

utils.py

@@ -112,7 +112,7 @@ def fetch_dataset_df(download_raw=False, test_repo=results_repo_validation): # C
         feature_schema = Features(raw_metric_features | other_features_raw)
         logger.info("Fetching raw bootstrapping dataset from Hugging Face Hub...")
         # Because the raw file is so long, we have to load it with delay and multiple retries
-        max_retries = 3
+        max_retries = 10
         base_delay = 5
         for attempt in range(max_retries):
             try: