push privacy fix

app.py

@@ -21,6 +21,7 @@ OWNER="gaia-benchmark"
 DATA_DATASET = f"{OWNER}/GAIA"
 INTERNAL_DATA_DATASET = f"{OWNER}/GAIA_internal"
 SUBMISSION_DATASET = f"{OWNER}/submissions_internal"
+CONTACT_DATASET = f"{OWNER}/contact_info"
 RESULTS_DATASET = f"{OWNER}/results_public"
 LEADERBOARD_PATH = f"{OWNER}/leaderboard"
 api = HfApi()
@@ -31,10 +32,11 @@ os.makedirs("scored", exist_ok=True)
 
 # Display the results
 eval_results = load_dataset(RESULTS_DATASET, YEAR_VERSION, token=TOKEN, download_mode="force_redownload", ignore_verifications=True)
+contact_infos = load_dataset(CONTACT_DATASET, YEAR_VERSION, token=TOKEN, download_mode="force_redownload", ignore_verifications=True)
 def get_dataframe_from_results(eval_results, split):
     local_df = eval_results[split]
     local_df = local_df.map(lambda row: {"model": model_hyperlink(row["url"], row["model"])})
-    local_df = local_df.remove_columns(["mail", "system_prompt", "url"])
+    local_df = local_df.remove_columns(["system_prompt", "url"])
     local_df = local_df.rename_column("model", "Model name")
     local_df = local_df.rename_column("model_family", "Model family")
     local_df = local_df.rename_column("score", "Average score (%)")
@@ -81,7 +83,7 @@ def add_new_eval(
     print("Adding new eval")
 
     # Check if the combination model/org already exists and prints a warning message if yes
-    if model.lower() in set(eval_results[val_or_test]["model"]) and organisation.lower() in set(eval_results[val_or_test]["organisation"]):
+    if model.lower() in set([m.lower() for m in eval_results[val_or_test]["model"]]) and organisation.lower() in set([o.lower() for l in eval_results[val_or_test]["organisation"]]):
         return format_warning("This model has been already submitted.")
     
     if path_to_file is None:
@@ -149,7 +151,6 @@ def add_new_eval(
         "system_prompt": system_prompt,
         "url": url,
         "organisation": organisation,
-        "mail": mail,
         "score": scores["all"]/num_questions["all"],
         "score_level1": scores[1]/num_questions[1],
         "score_level2": scores[2]/num_questions[2],
@@ -159,6 +160,16 @@ def add_new_eval(
     print(eval_results)
     eval_results.push_to_hub(RESULTS_DATASET, config_name = YEAR_VERSION, token=TOKEN)
 
+    contact_info = {
+        "model": model,
+        "model_family": model_family,
+        "url": url,
+        "organisation": organisation,
+        "mail": mail,
+    }
+    contact_infos[val_or_test]= contact_infos[val_or_test].add_item(contact_info)
+    contact_infos.push_to_hub(CONTACT_DATASET, config_name = YEAR_VERSION, token=TOKEN)
+
     return format_log(f"Model {model} submitted by {organisation} successfully. \nPlease refresh the leaderboard, and wait a bit to see the score displayed")
 
 
@@ -216,7 +227,7 @@ with demo:
                 url_textbox = gr.Textbox(label="Url to model information")
             with gr.Column():
                 organisation = gr.Textbox(label="Organisation")
-                mail = gr.Textbox(label="Contact email")
+                mail = gr.Textbox(label="Contact email (will be stored privately, & used if there is an issue with your submission)")
                 file_output = gr.File()