super basic leaderboard

app.py

@@ -7,9 +7,12 @@
 import gradio as gr
 from huggingface_hub import whoami
 from huggingface_hub.errors import LocalTokenNotFoundError
+from gradio_leaderboard import Leaderboard
 import argilla as rg
 import os
 from requests.exceptions import HTTPError
+import csv
+import pandas as pd
 
 CLA = """
                    Meta Platforms, Inc. Individual Contributor License Agreement ("Agreement"), v2.0
@@ -399,9 +402,19 @@ If you want to contribute dataset translations for a new language or validate ex
 
 
 def leaderboard_tab():
+    stats = pd.read_csv("data/benchmark_stats.tsv", sep="\t", quoting=csv.QUOTE_NONE)
+    df = stats.groupby(['system', 'level'])[
+        ['score_metricx_both', 'score_xcomet_both', 'score_CHRFpp', 'score_glotlid_ref']
+    ].mean().reset_index().sort_values('score_metricx_both')
     with gr.Tab("Leaderboard"):
-        gr.Markdown("Coming soon!")
-
+        gr.Markdown("The current leaderboard displays performance across all directions in the benchmark.")
+        gr.Markdown("A smarter leaderboard and the code for reproducing the evaluation will be published soon!")
+        # Leaderboard(
+        #     value=df,
+        #     select_columns=["system", "level"] +  ['score_metricx_both', 'score_xcomet_both', 'score_CHRFpp', 'score_glotlid_ref'],
+        #     filter_columns=["system", "level"],
+        # )
+        gr.Dataframe(df)
 
 with gr.Blocks(
     css="""
@@ -426,4 +439,3 @@ with gr.Blocks(
 
 if __name__ == "__main__":
     demo.launch()
-

requirements.txt

@@ -1,2 +1,4 @@
 argilla
 gradio[oauth]
+pandas
+gradio_leaderboard