Add search capability and language names

app.py

@@ -2,8 +2,10 @@ import os
 import json
 import glob
 from collections import defaultdict
+import pandas as pd
 import gradio as gr
 from content import *
+from css import *
 import glob
 
 ARC = "arc"
@@ -14,6 +16,42 @@ BENCHMARKS = [ARC, HELLASWAG, MMLU, TRUTHFULQA]
 
 METRICS = ["acc_norm", "acc_norm", "acc_norm", "mc2"]
 
+LANGS = 'ar,bn,ca,da,de,es,eu,fr,gu,hi,hr,hu,hy,id,it,kn,ml,mr,ne,nl,pt,ro,ru,sk,sr,sv,ta,te,uk,vi,zh'.split(',')
+
+LANG_NAME = {
+    'ar': 'Arabic',
+    'bn': 'Bengali',
+    'ca': 'Catalan',
+    'da': 'Danish',
+    'de': 'German',
+    'es': 'Spanish',
+    'eu': 'Basque',
+    'fr': 'French',
+    'gu': 'Gujarati',
+    'hi': 'Hindi',
+    'hr': 'Croatian',
+    'hu': 'Hungarian',
+    'hy': 'Armenian',
+    'id': 'Indonesian',
+    'it': 'Italian',
+    'kn': 'Kannada',
+    'ml': 'Malayalam',
+    'mr': 'Marathi',
+    'ne': 'Nepali',
+    'nl': 'Dutch',
+    'pt': 'Portuguese',
+    'ro': 'Romanian',
+    'ru': 'Russian',
+    'sk': 'Slovak',
+    'sr': 'Serbian',
+    'sv': 'Swedish',
+    'ta': 'Tamil',
+    'te': 'Telugu',
+    'uk': 'Ukrainian',
+    'vi': 'Vietnamese',
+    'zh': 'Chinese'
+}
+
 
 def collect_results():
     performance_dict = defaultdict(dict)
@@ -52,6 +90,7 @@ def collect_results():
 def get_leaderboard_df(performance_dict, pretrained_models):
     df = list()
     for (pretrained, lang), perfs in performance_dict.items():
+        lang_name = LANG_NAME[lang]
         arc_perf = perfs.get(ARC, 0.0)
         hellaswag_perf = perfs.get(HELLASWAG, 0.0)
         mmlu_perf = perfs.get(MMLU, 0.0)
@@ -60,26 +99,40 @@ def get_leaderboard_df(performance_dict, pretrained_models):
         if arc_perf * hellaswag_perf * mmlu_perf * truthfulqa_perf == 0:
             continue
         avg = round((arc_perf + hellaswag_perf + mmlu_perf + truthfulqa_perf) / 4, 1)
-        row = [pretrained, lang, avg, arc_perf, hellaswag_perf, mmlu_perf, truthfulqa_perf]
+        notes = ' '.join([pretrained, lang_name, lang])
+        row = [pretrained, lang_name, lang, avg, arc_perf, hellaswag_perf, mmlu_perf, truthfulqa_perf, notes]
         df.append(row)
+
+    df = pd.DataFrame.from_records(df, columns=COLS)
+    df = df.sort_values(by=[AVERAGE_COL], ascending=False)
+    df = df[COLS]
+
     return df
 
 
+def search_table(df, query):
+    filtered_df = df[df[NOTES_COL].str.contains(query, case=False)]
+    return filtered_df
+
+
+
 MODEL_COL = "Model"
 LANG_COL = "Language"
+CODE_COL = "Code"
 AVERAGE_COL = "Average"
 ARC_COL = "ARC (25-shot)"
 HELLASWAG_COL = "HellaSwag (10-shot)️"
 MMLU_COL = "MMLU (5-shot)"
 TRUTHFULQA_COL = "TruthfulQA (0-shot)"
+NOTES_COL = "Notes"  # For search only
 
-COLS = [MODEL_COL, LANG_COL, AVERAGE_COL, ARC_COL, HELLASWAG_COL, MMLU_COL, TRUTHFULQA_COL]
-TYPES = ["str", "str", "number", "number", "number", "number", "number"]
+COLS = [MODEL_COL, LANG_COL, CODE_COL, AVERAGE_COL, ARC_COL, HELLASWAG_COL, MMLU_COL, TRUTHFULQA_COL, NOTES_COL]
+TYPES = ["str", "str", "str", "number", "number", "number", "number", "number", "str"]
 
 args = collect_results()
-leaderboard_df = get_leaderboard_df(*args)
+original_df = get_leaderboard_df(*args)
 
-demo = gr.Blocks()
+demo = gr.Blocks(css=CUSTOM_CSS)
 with demo:
     gr.HTML(TITLE)
     gr.Markdown(INTRO_TEXT, elem_classes="markdown-text")
@@ -91,13 +144,24 @@ with demo:
         )
 
         leaderboard_table = gr.components.Dataframe(
-            value=leaderboard_df,
+            value=original_df,
             headers=COLS,
             datatype=TYPES,
             max_rows=5,
             elem_id="leaderboard-table",
         )
 
+        # # Dummy leaderboard for handling the case when the user uses backspace key
+        hidden_leaderboard_table_for_search = gr.components.Dataframe(
+            value=original_df, headers=COLS, datatype=TYPES, max_rows=5, visible=False
+        )
+
+        search_bar.change(
+            search_table,
+            [hidden_leaderboard_table_for_search, search_bar],
+            leaderboard_table,
+        )
+
     gr.Markdown(CREDIT, elem_classes="markdown-text")
     gr.Markdown(CITATION, elem_classes="markdown-text")
 

content.py

@@ -3,7 +3,7 @@ TITLE = '<h1 align="center" id="space-title">Open Multilingual LLM Evaluation Le
 INTRO_TEXT = f"""
 ## About
 
-This leaderboard shows the performance of pretrained models in 29 languages on four benchmarks:
+This leaderboard shows the performance of pretrained models in 29 languages including Arabic, Armenian, Basque, Bengali, Catalan, Chinese, Croatian, Danish, Dutch, French, German, Gujarati, Hindi, Hungarian, Indonesian, Italian, Kannada, Malayalam, Marathi, Nepali, Portuguese, Romanian, Russian, Serbian, Slovak, Spanish, Swedish, Tamil, Telugu, Ukrainian, and Vietnameseon four benchmarks:
 
 - <a href="https://arxiv.org/abs/1803.05457" target="_blank">  AI2 Reasoning Challenge </a> (25-shot) 
 - <a href="https://arxiv.org/abs/1905.07830" target="_blank">  HellaSwag </a> (10-shot) 

css.py

@@ -0,0 +1,13 @@
+CUSTOM_CSS= """
+/* Hides the final column */
+table td:last-child,
+table th:last-child {
+    display: none;
+}
+# table td:first-child,
+# table th:first-child {
+#     max-width: 400px;
+#     overflow: auto;
+#     white-space: nowrap;
+# }
+"""