Update app.py

app.py

@@ -1,7 +1,7 @@
 import gradio as gr
 import pandas as pd
 import json
-from constants import BANNER, INTRODUCTION_TEXT, CITATION_TEXT, METRICS_TAB_TEXT, DIR_OUTPUT_REQUESTS, LEADERBOARD_CSS, EXPLANATION
+from constants import LEADERBOARD_CSS, EXPLANATION, EXPLANATION_EDACC, EXPLANATION_AFRI
 from init import is_model_on_hub, upload_file, load_all_info_from_dataset_hub
 from utils_display import AutoEvalColumn, fields, make_clickable_model, styled_error, styled_message
 from datetime import datetime, timezone
@@ -48,6 +48,26 @@ column_names = {
     "Spanish_male": "Spanish male",
     "Vietnamese_female": "Vietnamese female",
     "Vietnamese_male": "Vietnamese male",
+    "agatu_test": "Agatu",
+    "angas_test": "Angas",
+    "bajju_test": "Bajju",
+    "bini_test": "Bini",
+    "brass_test": "Brass",
+    "delta_test": "Delta",
+    "eggon_test": "Eggon",
+    "ekene_test": "Ekene",
+    "ekpeye_test": "Ekpeye",
+    "gbagyi_test": "Gbagyi",
+    "igarra_test": "Igarra",
+    "ijaw-nembe_test": "Ijaw-Nembe",
+    "ikulu_test": "Ikulu",
+    "jaba_test": "Jaba",
+    "jukun_test": "Jukun",
+    "khana_test": "Khana",
+    "mada_test": "Mada",
+    "mwaghavul_test": "Mwaghavul",
+    "ukwuani_test": "Ukwuani",
+    "yoruba-hausa_test": "Yoruba-Hausa",
 }
 
 african_cols = ["Ghanain English female", "Kenyan English female", "Kenyan English male", "Nigerian English female", "Nigerian English male"]
@@ -57,13 +77,15 @@ latin_american_cols = ["Latin American female", "Latin American male"]
 british_cols = ["Irish English female", "Irish English male", "Scottish English male", "Southern British English male"]
 european_cols = ["Eastern European male", "European male", "French female", "Italian female", "Spanish female", "Spanish male", "Catalan female", "Bulgarian female", "Bulgarian male", "Lithuanian male", "Romanian female"]
 asian_cols = ["Chinese female", "Chinese male", "Indonesian female", "Vietnamese female", "Vietnamese male", "Indian English female", "Indian English male"]
-eval_queue_repo, requested_models, csv_results = load_all_info_from_dataset_hub()
+eval_queue_repo_edacc, requested_models, csv_results_edacc, csv_results_afrispeech = load_all_info_from_dataset_hub()
 
-if not csv_results.exists():
-    raise Exception(f"CSV file {csv_results} does not exist locally")
+if not csv_results_edacc.exists():
+    raise Exception(f"CSV file {csv_results_edacc} does not exist locally")
     
 # Get csv with data and parse columns
-original_df = pd.read_csv(csv_results)
+original_df = pd.read_csv(csv_results_edacc)
+
+afrispeech_df = pd.read_csv(csv_results_afrispeech)
 
 # Formats the columns
 def formatter(x):
@@ -79,8 +101,17 @@ for col in original_df.columns:
     else:
         original_df[col] = original_df[col].apply(formatter) # For numerical values
  
+for col in afrispeech_df.columns:
+    if col == "model":
+        afrispeech_df[col] = afrispeech_df[col].apply(lambda x: x.replace(x, make_clickable_model(x)))
+    else:
+        afrispeech_df[col] = afrispeech_df[col].apply(formatter) # For numerical values
+
 original_df.rename(columns=column_names, inplace=True)
 original_df.sort_values(by='Average WER ⬇️', inplace=True)
+afrispeech_df.rename(columns=column_names, inplace=True)
+afrispeech_df.sort_values(by='Average WER ⬇️', inplace=True)
+
 female_cols = [col for col in original_df.columns if 'female' == col.split(' ')[-1]]
 male_cols = [col for col in original_df.columns if 'male' == col.split(' ')[-1]]
 
@@ -150,13 +181,14 @@ TYPES = [c.type for c in fields(AutoEvalColumn)]
 with gr.Blocks(css=LEADERBOARD_CSS) as demo:
     # gr.HTML(BANNER, elem_id="banner")
     # Write a header with the title
-    gr.Markdown("<h1>🤗 Testing Model Robustness on EdAcc</h1>", elem_classes="markdown-text")
+    gr.Markdown("<h1>🤫 How Biased is Whisper?</h1>", elem_classes="markdown-text")
 
     
     gr.Markdown(EXPLANATION, elem_classes="markdown-text")
 
     with gr.Tabs(elem_classes="tab-buttons") as tabs:
-        with gr.TabItem("🏅 Leaderboard", elem_id="od-benchmark-tab-table", id=0):
+        with gr.TabItem("🏅 Edacc Results", elem_id="od-benchmark-tab-table", id=0):
+            gr.Markdown(EXPLANATION_EDACC, elem_classes="markdown-text")
             # Add column filter dropdown
             column_filter = gr.Dropdown(
                 choices=["All", "Female", "Male", "African", "North American", "Caribbean", "Latin American", "British", "European", "Asian"] + [v for k,v in column_names.items() if k != "model"],
@@ -217,4 +249,37 @@ with gr.Blocks(css=LEADERBOARD_CSS) as demo:
                 outputs=[leaderboard_table]
             )
 
+        with gr.TabItem("🏅 Afrispeech Results", elem_id="od-benchmark-tab-table", id=1):
+            gr.Markdown(EXPLANATION_AFRI, elem_classes="markdown-text")
+            # Add column filter dropdown
+            afrispeech_column_filter = gr.Dropdown(
+                choices=["All"] + [v for k,v in column_names.items() if k != "model" and v in afrispeech_df.columns],
+                label="Filter by column", 
+                multiselect=True,
+                value=["All"],
+                elem_id="afrispeech-column-filter"
+            )
+            
+            leaderboard_table = gr.components.Dataframe(
+                value=afrispeech_df,
+                datatype=TYPES,
+                elem_id="leaderboard-table",
+                interactive=False,
+                visible=True,
+            )
+
+            # Update table columns when filter changes
+            def update_afrispeech_table(cols):
+                if "All" in cols:
+                    return gr.Dataframe(value=afrispeech_df)
+                
+                selected_cols = ["Model"] + cols  # Always include the Model column
+                return gr.Dataframe(value=afrispeech_df[selected_cols])
+            
+            afrispeech_column_filter.change(
+                fn=update_afrispeech_table,
+                inputs=[afrispeech_column_filter],
+                outputs=[leaderboard_table]
+            )
+
 demo.launch(ssr_mode=False)