moved all constants + added status for uploaded tresults

app.py

@@ -6,51 +6,12 @@ from utils import (
     create_html_image,
 )
 import os
-from datasets import load_dataset
 import gradio as gr
 
+from constants import *
+
 if __name__ == "__main__":
     
-    # Evaluation dataset path
-    DATA_PATH = "atlasia/Arabic-LID-Leaderboard"
-    
-    # All Metrics
-    metrics = [
-        'f1_score',
-        'precision',
-        'recall',
-        'specificity',
-        'false_positive_rate',
-        'false_negative_rate',
-        'negative_predictive_value',
-        'n_test_samples',
-    ]
-    # Default metrics to display    
-    default_metrics = [
-        'f1_score', 
-        'precision', 
-        'recall', 
-        'false_positive_rate', 
-        'false_negative_rate'
-    ]
-    # default language to display
-    default_languages = [
-        'Morocco',
-        'MSA',
-        'Egypt',
-        'Algeria',
-        'Tunisia',
-        'Levantine',
-    ]
-    
-    # Load test dataset
-    test_dataset = load_dataset(DATA_PATH, split='test')   
-    
-    # Supported dialects
-    all_target_languages = list(test_dataset.unique("dialect"))
-    supported_dialects = all_target_languages + ['All']
-    languages_to_display_one_vs_all = all_target_languages # everything except All
-        
     with gr.Blocks() as app:
         base_path = os.path.dirname(__file__)
         local_image_path = os.path.join(base_path, 'open_arabic_lid_arena.png')
@@ -58,8 +19,8 @@ if __name__ == "__main__":
         gr.HTML(create_html_image(local_image_path))
         gr.Markdown("# 🏅 Open Arabic Dialect Identification Leaderboard")
         
-        # Multilingual model leaderboard
-        with gr.Tab("Multilingual model leaderboard"):
+        # Multi-dialects leaderboard
+        with gr.Tab("Multi-dialects model leaderboard"):
             gr.Markdown("""
                         Complete leaderboard across multiple arabic dialects. 
                         Compare the performance of different models across various metrics such as FNR, FPR, and other clasical metrics.
@@ -97,13 +58,13 @@ if __name__ == "__main__":
                         Let's work together to improve Arabic dialect identification! 🚀
                         """)
             
-        # Binary model leaderboard
-        with gr.Tab("One-vs-All leaderboard"):
+        # Dialect confusion leaderboard
+        with gr.Tab("Dialect confusion leaderboard"): # use to be "One-vs-All leaderboard"
             
             gr.Markdown("""
-                        A kind of one-vs-all approach for evaluating LID models across multiple arabic dialects. 
-                        Computes the `false_positive_rate` of different models for a given target language.
-                        This should help you understand how well a model can identify a specific dialect by
+                        Detailed analysis of how well models distinguish specific dialects from others.
+                        For each target dialect, see how often models incorrectly classify other dialects as the target.
+                        Lower `false_positive_rate` indicate better ability to identify the true dialect by
                         showing **how often it misclassifies other dialects as the target dialect**.
                         """
             )
@@ -138,8 +99,8 @@ if __name__ == "__main__":
             model_path = gr.Textbox(label="Model Path", placeholder='path/to/model')
             model_path_bin = gr.Textbox(label=".bin filename", placeholder='model.bin')
             gr.Markdown("### **⚠️ To ensure correct results, tick this when the model's labels are the iso_codes**")
-            use_mapping = gr.Checkbox(label="Does not map to country")
-            eval_button = gr.Button("Evaluate", value=False)  # Initially disabled
+            use_mapping = gr.Checkbox(label="Does not map to country", value=True)      # Initially enabled
+            eval_button = gr.Button("Evaluate", value=False)                            # Initially disabled
             
             # Status message area
             status_message = gr.Markdown(value="")
@@ -160,7 +121,7 @@ if __name__ == "__main__":
             model = ... # Load your model here
             
             # Load evaluation benchmark
-            eval_dataset = load_dataset("atlasia/No-Arabic-Dialect-Left-Behind-Filtered-Balanced", split='test').to_pandas() # do not change this line :)
+            eval_dataset = load_dataset("atlasia/Arabic-LID-Leaderboard", split='test').to_pandas() # do not change this line :)
             
             # Predict labels using your model
             eval_dataset['preds'] = eval_dataset['text'].apply(lambda text: predict_label(text, model)) # predict_label is a function that you need to define for your model
@@ -182,7 +143,15 @@ if __name__ == "__main__":
             uploaded_model_name = gr.Textbox(label="Model name", placeholder='Your model/team name')
             file = gr.File(label="Upload your results")
             upload_button = gr.Button("Upload")
-            upload_button.click(process_results_file, inputs=[file, uploaded_model_name], outputs=[leaderboard_table])
+            
+            # Status message area
+            status_message = gr.Markdown(value="")
+            def update_status_message():
+                return "### **⚠️Evaluating... Please wait...**"
+            
+            upload_button.click(update_status_message, outputs=[status_message])
+            
+            upload_button.click(process_results_file, inputs=[file, uploaded_model_name], outputs=[leaderboard_table, status_message])
         
         # Update multilangual table when any input changes
         country_selector.change(