cleanup

app.py

@@ -9,17 +9,11 @@ import numpy as np
 import pandas as pd
 
 
-# notebook_url = "https://colab.research.google.com/drive/1RAWb22-PFNI-X1gPVzc927SGUdfr6nsR?usp=sharing"
-notebook_url = "https://colab.research.google.com/drive/1KdwokPjirkTmpO_P1WByFNFiqxWQquwH#scrollTo=o_CpbkGEbhrK"
-
-
-basic_component_values = [None] * 6
-leader_component_values = [None] * 5
 
 
 def make_default_md():
     leaderboard_md = f"""
-# 🏆 BabilongLeaderboard
+# 🏆 Babilong Leaderboard
 | [GitHub](https://github.com/lm-sys/FastChat) | [Paper](https://arxiv.org/abs/2306.05685) | [Dataset](https://github.com/lm-sys/FastChat/blob/main/docs/dataset_release.md) |
 """
     return leaderboard_md
@@ -31,80 +25,6 @@ def make_arena_leaderboard_md():
     return leaderboard_md
 
 
-def make_full_leaderboard_md(elo_results):
-    leaderboard_md = f"""
-Three benchmarks are displayed: **Arena Elo**, **MT-Bench** and **MMLU**.
-- [Chatbot Arena](https://chat.lmsys.org/?arena) - a crowdsourced, randomized battle platform. We use 200K+ user votes to compute Elo ratings.
-- [MT-Bench](https://arxiv.org/abs/2306.05685): a set of challenging multi-turn questions. We use GPT-4 to grade the model responses.
-- [MMLU](https://arxiv.org/abs/2009.03300) (5-shot): a test to measure a model's multitask accuracy on 57 tasks.
-
-💻 Code: The MT-bench scores (single-answer grading on a scale of 10) are computed by [fastchat.llm_judge](https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge).
-The MMLU scores are mostly computed by [InstructEval](https://github.com/declare-lab/instruct-eval).
-Higher values are better for all benchmarks. Empty cells mean not available.
-"""
-    return leaderboard_md
-
-
-def make_leaderboard_md_live(elo_results):
-    leaderboard_md = f"""
-# Leaderboard
-Last updated: {elo_results["last_updated_datetime"]}
-{elo_results["leaderboard_table"]}
-"""
-    return leaderboard_md
-
-
-def update_elo_components(max_num_files, elo_results_file):
-    log_files = get_log_files(max_num_files)
-
-    # Leaderboard
-    if elo_results_file is None:  # Do live update
-        battles = clean_battle_data(log_files)
-        elo_results = report_elo_analysis_results(battles)
-
-        leader_component_values[0] = make_leaderboard_md_live(elo_results)
-        leader_component_values[1] = elo_results["win_fraction_heatmap"]
-        leader_component_values[2] = elo_results["battle_count_heatmap"]
-        leader_component_values[3] = elo_results["bootstrap_elo_rating"]
-        leader_component_values[4] = elo_results["average_win_rate_bar"]
-
-    # Basic stats
-    basic_stats = report_basic_stats(log_files)
-    md0 = f"Last updated: {basic_stats['last_updated_datetime']}"
-
-    md1 = "### Action Histogram\n"
-    md1 += basic_stats["action_hist_md"] + "\n"
-
-    md2 = "### Anony. Vote Histogram\n"
-    md2 += basic_stats["anony_vote_hist_md"] + "\n"
-
-    md3 = "### Model Call Histogram\n"
-    md3 += basic_stats["model_hist_md"] + "\n"
-
-    md4 = "### Model Call (Last 24 Hours)\n"
-    md4 += basic_stats["num_chats_last_24_hours"] + "\n"
-
-    basic_component_values[0] = md0
-    basic_component_values[1] = basic_stats["chat_dates_bar"]
-    basic_component_values[2] = md1
-    basic_component_values[3] = md2
-    basic_component_values[4] = md3
-    basic_component_values[5] = md4
-
-
-def update_worker(max_num_files, interval, elo_results_file):
-    while True:
-        tic = time.time()
-        update_elo_components(max_num_files, elo_results_file)
-        durtaion = time.time() - tic
-        print(f"update duration: {durtaion:.2f} s")
-        time.sleep(max(interval - durtaion, 0))
-
-
-def load_demo(url_params, request: gr.Request):
-    logger.info(f"load_demo. ip: {request.client.host}. params: {url_params}")
-    return basic_component_values + leader_component_values
-
 
 def model_hyperlink(model_name, link):
     return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'
@@ -147,48 +67,6 @@ def load_leaderboard_table_csv(filename, add_hyperlink=True):
     return rows
 
 
-def build_basic_stats_tab():
-    empty = "Loading ..."
-    basic_component_values[:] = [empty, None, empty, empty, empty, empty]
-
-    md0 = gr.Markdown(empty)
-    gr.Markdown("#### Figure 1: Number of model calls and votes")
-    plot_1 = gr.Plot(show_label=False)
-    with gr.Row():
-        with gr.Column():
-            md1 = gr.Markdown(empty)
-        with gr.Column():
-            md2 = gr.Markdown(empty)
-    with gr.Row():
-        with gr.Column():
-            md3 = gr.Markdown(empty)
-        with gr.Column():
-            md4 = gr.Markdown(empty)
-    return [md0, plot_1, md1, md2, md3, md4]
-
-def get_full_table(arena_df, model_table_df):
-    values = []
-    for i in range(len(model_table_df)):
-        row = []
-        model_key = model_table_df.iloc[i]["key"]
-        model_name = model_table_df.iloc[i]["Model"]
-        # model display name
-        row.append(model_name)
-        if model_key in arena_df.index:
-            idx = arena_df.index.get_loc(model_key)
-            row.append(round(arena_df.iloc[idx]["rating"]))
-        else:
-            row.append(np.nan)
-        row.append(model_table_df.iloc[i]["MT-bench (score)"])
-        row.append(model_table_df.iloc[i]["MMLU"])
-        # Organization
-        row.append(model_table_df.iloc[i]["Organization"])
-        # license
-        row.append(model_table_df.iloc[i]["License"])
-
-        values.append(row)
-    values.sort(key=lambda x: -x[1] if not np.isnan(x[1]) else 1e9)
-    return values
     
 
 def build_leaderboard_tab():
@@ -204,27 +82,25 @@ def build_leaderboard_tab():
                 headers=[
                     "Rank",
                     "🤖 Model",
-                    "⭐ Arena Elo",
-                    "📊 95% CI",
-                    "🗳️ Votes",
-                    "Organization",
-                    "License",
-                    "Knowledge Cutoff",
+                    "qa 1",
+                    "qa 2",
+                    "qa 3",
+                    "qa 4",
+                    "qa 5",
                 ],
                 datatype=[
                     "str",
                     "markdown",
                     "number",
-                    "str",
                     "number",
-                    "str",
-                    "str",
-                    "str",
+                    "number",
+                    "number",
+                    "number",
                 ],
                 # value=arena_table_vals,
                 elem_id="arena_leaderboard_dataframe",
                 height=700,
-                column_widths=[50, 200, 120, 100, 100, 150, 150, 100],
+                column_widths=[50, 200, 150, 150, 150, 150, 150],
                 wrap=True,
             )
    
@@ -268,17 +144,7 @@ footer {
 }
 """
 
-acknowledgment_md = """
-### Acknowledgment
-<div class="image-container">
-    <p> We thank <a href="https://www.kaggle.com/" target="_blank">Kaggle</a>, <a href="https://mbzuai.ac.ae/" target="_blank">MBZUAI</a>, <a href="https://www.anyscale.com/" target="_blank">AnyScale</a>, <a href="https://www.a16z.com/" target="_blank">a16z</a>, and <a href="https://huggingface.co/" target="_blank">HuggingFace</a> for their generous <a href="https://lmsys.org/donations/" target="_blank">sponsorship</a>. </p>
-    <img src="https://upload.wikimedia.org/wikipedia/commons/thumb/7/7c/Kaggle_logo.png/400px-Kaggle_logo.png" alt="Kaggle">
-    <img src="https://mma.prnewswire.com/media/1227419/MBZUAI_Logo.jpg?p=facebookg" alt="MBZUAI">
-    <img src="https://docs.anyscale.com/site-assets/logo.png" alt="AnyScale">
-    <img src="https://a16z.com/wp-content/themes/a16z/assets/images/opegraph_images/corporate-Yoast-Twitter.jpg" alt="a16z">
-    <img src="https://huggingface.co/datasets/huggingface/brand-assets/resolve/main/hf-logo-with-title.png" alt="HuggingFace">
-</div>
-"""
+
 
 def build_demo():
     text_size = gr.themes.sizes.text_lg