support dark mode

app.py

@@ -7,18 +7,17 @@ from dotenv import load_dotenv
 from matplotlib.colors import LinearSegmentedColormap
 import plotly.express as px
 import plotly.graph_objects as go
-from sklearn.linear_model import LinearRegression
+# from sklearn.linear_model import LinearRegression
 import numpy as np
 from huggingface_hub import HfApi
 from huggingface_hub.hf_api import HTTPError
 from huggingface_hub.utils import GatedRepoError
 from gradio_rangeslider import RangeSlider
 import datetime
-
+from gradio.themes.utils.colors import slate
 
 load_dotenv()
 webhook_url = os.environ.get("WEBHOOK_URL")
-
 file_name_list = [
     "14b",
     "9b",
@@ -27,19 +26,16 @@ file_name_list = [
     "1b5",
     "other",
 ]
-
 sheet_name_list = [
     "cr",
     "bpc",
     "bpb",
 ]
-
 metric_list = [
     "Compression Rate (%)",
     "Bits Per Character (BPC)",
     "Bits Per Byte (BPB)",
 ]
-
 model_size_list = [
     "~14B",
     "~9B",
@@ -48,13 +44,11 @@ model_size_list = [
     "~1.5B",
     "Other",
 ]
-
 metric_to_sheet = {
     "Compression Rate (%)": "cr",
     "Bits Per Character (BPC)": "bpc",
     "Bits Per Byte (BPB)": "bpb",
 }
-
 model_size_to_file_name = {
     "~14B": "14b",
     "~9B": "9b",
@@ -68,27 +62,21 @@ def read_about_md():
     with open('about.md', 'r', encoding='utf-8') as f:
         return f.read()
 
-
 def rename_columns(df):
     df.columns = [col.rsplit("_", maxsplit=1)[0] for col in df.columns]
     return df
 
-
 def get_folders_matching_format(directory):
     pattern = re.compile(r"^\d{4}-\d{2}$")
     folders = []
-
     if not os.path.exists(directory):
         return folders
-
     for item in os.listdir(directory):
         full_path = os.path.join(directory, item)
         if os.path.isdir(full_path) and pattern.match(item):
             folders.append(full_path)
-
     return folders
 
-
 def get_unique_column_names(data=None):
     return [
         "ao3_\u200benglish",
@@ -100,74 +88,63 @@ def get_unique_column_names(data=None):
         "github_\u200bpython",
     ]
 
-
 def color_cell(value):
     return "background-color: #fffdd0" if pd.notna(value) else "default"
 
+# def color_cell_themed(value):
+#     return "background-color: rgba(255, 253, 208, 1.0)" if pd.notna(value) else "default"
 
-def update_table(
-    period: str,
-    models_size: list,
-    metric: str,
-    visible_columns: list,
-    color_columns: list,
-    size_range: list,
-    midpoint: float = 0.5,
-    sort_by: str = "Average (lower=better)",
-    ascending: bool = True,
-):
-    print(
-        f"Updating - time: {datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')}, period: {period}, models: {models_size}, metric: {metric}, visible_columns: {visible_columns}, color_columns: {color_columns}, size_range: {size_range}, sort_by: {sort_by}, ascending: {ascending}\n"
-    )
+# --- 核心改动点 1: 修改 update_table 函数 ---
+# 添加 request: gr.Request = None 参数来接收主题模式信息
+# 默认值为 None 是为了处理初始加载
+def update_table(period: str, models_size: list, metric: str, visible_columns: list, color_columns: list, size_range: list, midpoint: float = 0.5, sort_by: str = "Average (lower=better)", ascending: bool = True, request: gr.Request = None):
+    # 打印日志并检查当前模式
+    is_dark_mode = request.is_dark if request else False
+    print(f"Updating - time: {datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')}, period: {period}, models: {models_size}, metric: {metric}, visible_columns: {visible_columns}, color_columns: {color_columns}, size_range: {size_range}, sort_by: {sort_by}, ascending: {ascending}, is_dark: {is_dark_mode}\n")
 
     if not models_size:
         return "No data available for the selected models and period."
-        # return pd.DataFrame()
-
+        
     target_period_data = all_data[period]
     target_file_name = [model_size_to_file_name[model] for model in models_size]
     sheet_name = metric_to_sheet[metric]
-
-    # combined_data = pd.concat([target_period_data[file_name][sheet_name] for file_name in target_file_name], axis=0)
-    combined_data = pd.concat(
-        [df.dropna(axis=1, how="all") for df in [target_period_data[file_name][sheet_name] for file_name in target_file_name]], axis=0
-    )
+    combined_data = pd.concat([df.dropna(axis=1, how="all") for df in [target_period_data[file_name][sheet_name] for file_name in target_file_name]], axis=0)
+    
     if len(combined_data) == 0:
         return "No data available for the selected models and period."
-        # return pd.DataFrame()
-
-    # Filter models based on the size range
+        
     combined_data = combined_data[combined_data["Parameters Count (B)"].between(size_range[0], size_range[1])]
     combined_data.reset_index(drop=True, inplace=True)
+    
     if len(combined_data) == 0:
         return "No data available for the selected models and period."
-        # return pd.DataFrame()
-
+        
     combined_data["Name"] = combined_data["Name"].apply(lambda x: x.replace(".pth", ""))
-
     ordered_columns = get_unique_column_names()
     relevant_columns = [col for col in ordered_columns if col in visible_columns and col not in ["Name", "Parameters Count (B)", "Average (The lower the better)"]]
     
-    if len(combined_data) > 0:
+    if len(combined_data) > 0 and relevant_columns:
         combined_data["Average (The lower the better)"] = round(combined_data[relevant_columns].mean(axis=1), 3)
-    combined_data = combined_data.rename(columns={"Parameters Count (B)": "Params (B)"})
-    combined_data = combined_data.rename(columns={"Average (The lower the better)": "Average (lower=better)"})
+        
+    combined_data = combined_data.rename(columns={"Parameters Count (B)": "Params (B)", "Average (The lower the better)": "Average (lower=better)"})
     sorted_data = combined_data.sort_values(by=sort_by, ascending=ascending)
-    visible_columns = ["Name", "Params (B)", "Average (lower=better)"] + relevant_columns
-    filtered_data = sorted_data[visible_columns]
+    visible_columns_final = ["Name", "Params (B)", "Average (lower=better)"] + relevant_columns
+    filtered_data = sorted_data[visible_columns_final]
     filtered_data.columns = [col.replace("_", " ") for col in filtered_data.columns]
-
     formatter = {col: "{:.3f}" for col in filtered_data.columns if filtered_data[col].dtype in ["float64", "float32"]}
-
-    # color gradient
-    colors = ["#63be7b", "#ffffff", "#f8696b"]
-    vmin = {}
-    vmax = {}
-    vmid = {}
+    
+    # --- 核心改动点 2: 根据主题模式选择不同的配色方案 ---
+    if is_dark_mode:
+        # 夜间模式配色 (绿 -> 深灰 -> 红)
+        colors = ["#2ca02c", "#2b2b2b", "#d62728"]
+    else:
+        # 日间模式配色 (绿 -> 白 -> 红)
+        colors = ["#63be7b", "#ffffff", "#f8696b"]
+    
+    vmin, vmax, vmid = {}, {}, {}
     for column in filtered_data.columns:
-        if column in ["Name", "Params (B)"]:
-            continue
-        col_values = filtered_data[column]
+        if column in ["Name", "Params (B)"]: continue
+        col_values = filtered_data[column].dropna()
         if len(col_values) > 1:
             sorted_values = np.sort(col_values)
             vmin[column] = sorted_values.min()
@@ -175,99 +152,64 @@ def update_table(
             idx = int(len(sorted_values) * midpoint)
             vmid[column] = sorted_values[idx]
 
-    def custom_background_gradient(series, cmap, vmin, vmax, vmid):
-        if len(series) == 0:
-            return series
-
+    # --- 核心改动点 3: 修改样式函数以包含固定的黑色字体 ---
+    def custom_background_gradient(series, cmap, vmin_val, vmax_val, vmid_val):
+        if len(series) == 0: return series
         def normalize(x):
-            if x <= vmid:
-                return 0.5 * (x - vmin) / (vmid - vmin)
+            if pd.isna(x): return 0.5 # Neutral for NaN
+            if vmid_val == vmin_val and x <= vmid_val: return 0.0
+            if vmid_val == vmax_val and x >= vmid_val: return 1.0
+            if vmid_val == vmin_val or vmid_val == vmax_val: return 0.5
+            if x <= vmid_val:
+                return 0.5 * (x - vmin_val) / (vmid_val - vmin_val)
             else:
-                return 0.5 + 0.5 * (x - vmid) / (vmax - vmid)
-
+                return 0.5 + 0.5 * (x - vmid_val) / (vmax_val - vmid_val)
         normed = series.apply(normalize)
-        colors = [cmap(x) for x in normed]
-        return ["background-color: rgba({}, {}, {}, {})".format(*[int(255 * x) for x in c[:3]], c[3]) for c in colors]
+        cmap_colors = [cmap(x) for x in normed]
+        # 在返回的CSS中同时设置 background-color 和 color
+        return [
+            "background-color: rgba({}, {}, {}, {}); color: black;".format(*[int(255 * c) for c in color[:3]], color[3])
+            for color in cmap_colors
+        ]
 
     target_color_columns = []
-    if "Average" in color_columns:
-        target_color_columns.append("Average (lower=better)")
-    if "Individual Tests" in color_columns:
-        target_color_columns.extend([col for col in filtered_data.columns if col not in ["Name", "Params (B)", "Average (lower=better)"]])
-
-    styler = filtered_data.style.format(formatter).map(color_cell, subset=["Params (B)"])
-
+    if "Average" in color_columns: target_color_columns.append("Average (lower=better)")
+    if "Individual Tests" in color_columns: target_color_columns.extend([col for col in filtered_data.columns if col not in ["Name", "Params (B)", "Average (lower=better)"]])
+    
+    def color_params_column_dynamic(value):
+        if not pd.notna(value):
+            return "default"
+        
+        # 2. 根据 is_dark_mode 返回不同的颜色
+        if is_dark_mode:
+            # 为夜间模式选择一个柔和、不刺眼的暗金色
+            # 字体颜色也设置为浅色以保证对比度
+            return "background-color: #4b4936; color: #f0f0f0;"
+        else:
+            # 为日间模式使用明亮的奶油色，字体为黑色
+            return "background-color: #fffdd0; color: black;"
+    
+    styler = filtered_data.style.format(formatter).map(color_params_column_dynamic, subset=["Params (B)"])
     for column in target_color_columns:
-        styler = styler.apply(
-            custom_background_gradient,
-            cmap=LinearSegmentedColormap.from_list("custom_cmap", colors),
-            vmin=vmin[column],
-            vmax=vmax[column],
-            vmid=vmid[column],
-            subset=[column],
-        )
-
-    # return styler
+        if column in vmin:
+            custom_cmap = LinearSegmentedColormap.from_list("custom_cmap", colors)
+            styler = styler.apply(custom_background_gradient, cmap=custom_cmap, vmin_val=vmin[column], vmax_val=vmax[column], vmid_val=vmid[column], subset=[column])
+          
     styler = styler.hide(axis="index")
-
     widths = [300, 150, 150, 100, 100, 100, 100, 100, 100, 100, 100]
+    
     table_styles = []
-
+    table_styles.append({"selector": "th", "props": [("background-color", "var(--background-fill-secondary)"), ("color", "var(--body-text-color)"), ("padding", "8px"), ("font-weight", "bold")]})
+    table_styles.append({"selector": "table", "props": [("border-collapse", "collapse"), ("border", f"1px solid var(--border-color-primary)")]})
     for i, w in enumerate(widths):
-        table_styles.append(
-            {
-                "selector": "th",
-                "props": [
-                    ("background-color", "#f5f5f5"),
-                    ("padding", "8px"),
-                    ("font-weight", "bold"),
-                ],
-            }
-        )
-        table_styles.append(
-            {
-                "selector": f"th.col{i}",
-                "props": [
-                    ("min-width", f"{w}px"),
-                    ("max-width", f"{w}px"),
-                    ("text-align", "center"),
-                    ("border", "1px solid #dddddd"),
-                ],
-            }
-        )
-        table_styles.append(
-            {
-                "selector": f"td.col{i}",
-                "props": [
-                    ("min-width", f"{w}px"),
-                    ("max-width", f"{w}px"),
-                    ("text-align", "center"),
-                    ("border", "1px solid #dddddd"),
-                ],
-            }
-        )
-
-    table_styles.append(
-        {
-            "selector": "table",
-            "props": [
-                ("border-collapse", "collapse"),
-                ("border", "1px solid #dddddd"),
-            ],
-        }
-    )
-
+        table_styles.append({"selector": f"th.col{i}, td.col{i}", "props": [("min-width", f"{w}px"), ("max-width", f"{w}px"), ("text-align", "center"), ("border", f"1px solid var(--border-color-primary)")]})
     styler = styler.set_table_styles(table_styles)
-
-    html_output = styler.to_html()
-    return html_output
-
+    return styler.to_html()
 
 def create_world_languages_gdp_chart():
     languages = ["English", "Chinese", "Spanish", "Japanese", "German", "French", "Arabic", "Italian", "Portuguese", "Korean", "Other"]
     shares = [27, 18, 8, 6, 5, 4, 3, 2, 2, 2, 23]
     colors = ["#FF7F7F", "#FFA07A", "#FFDB58", "#90EE90", "#98FB98", "#87CEFA", "#B0C4DE", "#DDA0DD", "#D8BFD8", "#F0E68C", "#E0FFFF"]
-
     fig = go.Figure(
         data=[
             go.Pie(
@@ -282,7 +224,6 @@ def create_world_languages_gdp_chart():
             )
         ]
     )
-
     fig.update_layout(
         title={
             "text": "World Languages by Share of Global GDP",
@@ -297,10 +238,8 @@ def create_world_languages_gdp_chart():
         height=500,
         margin=dict(t=80, b=20, l=20, r=20),
     )
-
     return fig
 
-
 def check_model_exists(model_id):
     api = HfApi()
     try:
@@ -314,11 +253,9 @@ def check_model_exists(model_id):
         else:
             return "Error: " + str(e)
 
-
 def submit_model(name):
     if "Exists" not in check_model_exists(name):
         return f"# ERROR: Model {name} does not exist on Hugging Face!"
-
     try:
         response = requests.post(webhook_url, json={"content": name})
         if response.status_code == 200:
@@ -334,131 +271,20 @@ def submit_model(name):
     except Exception as e:
         print(e)
         return "ERROR: Unexpected error. Please try again later."
-
-
-# def create_scaling_plot(all_data, period):
-#     selected_columns = ["Name", "Parameters Count (B)", "Average (The lower the better)"]
-#     target_data = all_data[period]
-#     new_df = pd.DataFrame()
-
-#     for size in target_data.keys():
-#         new_df = pd.concat([new_df, target_data[size]["cr"].loc[:, selected_columns].dropna(axis=1, how="all")], axis=0)
-
-#     new_df.rename(columns={"Parameters Count (B)": "Params(B)", "Average (The lower the better)": "Compression Rate (%)"}, inplace=True)
-
-#     new_df["Log Params(B)"] = np.log(new_df["Params(B)"])
-#     new_df["Log Compression Rate (%)"] = np.log(new_df["Compression Rate (%)"])
-
-#     fig = px.scatter(
-#         new_df,
-#         x="Log Params(B)",
-#         y="Log Compression Rate (%)",
-#         title="Compression Rate Scaling Law",
-#         hover_name="Name",
-#         custom_data=["Params(B)", "Compression Rate (%)"],
-#     )
-
-#     fig.update_traces(
-#         hovertemplate="<b>%{hovertext}</b><br>Params(B): %{customdata[0]:.2f} B<br>Compression Rate (%): %{customdata[1]:.2f}<extra></extra>"
-#     )
-#     fig.update_layout(
-#         width=800,  # 设置图像宽度
-#         height=600,  # 设置图像高度
-#         title={"text": "Compression Rate Scaling Law", "x": 0.5, "xanchor": "center", "yanchor": "top"},
-#         showlegend=True,
-#         xaxis={"showgrid": True, "zeroline": False, "type": "linear", "title": "Params(B)"},  # 确保坐标轴类型正确
-#         yaxis={"showgrid": True, "zeroline": False, "type": "linear", "title": "Compression Rate (%)", "autorange": "reversed"},
-#     )
-
-#     names_to_connect_dict = {
-#         "2024-05": ["Meta-Llama-3-8B", "stablelm-3b-4e1t", "Qwen2-1.5B", "TinyLlama-1.1B-intermediate-step-1431k-3T", "Mistral-Nemo-Base-2407"],
-#         "2024-06": ["Meta-Llama-3-8B", "stablelm-3b-4e1t", "Qwen2-1.5B", "TinyLlama-1.1B-intermediate-step-1431k-3T", "Mistral-Nemo-Base-2407"],
-#         "2024-07": ["Meta-Llama-3.1-8B", "stablelm-3b-4e1t", "Qwen2-1.5B", "TinyLlama-1.1B-intermediate-step-1431k-3T", "Mistral-Nemo-Base-2407"],
-#         "2024-08": [
-#             "Meta-Llama-3.1-8B",
-#             "Rene-v0.1-1.3b-pytorch",
-#             "stablelm-3b-4e1t",
-#             "Qwen2-1.5B",
-#             "TinyLlama-1.1B-intermediate-step-1431k-3T",
-#             "Mistral-Nemo-Base-2407",
-#         ],
-#         "2025-01": ["Qwen2.5-1.5B"],
-#     }
-
-#     names_to_connect = names_to_connect_dict.get(period, names_to_connect_dict["2024-08"])
-
-#     connection_points = new_df[new_df["Name"].isin(names_to_connect)]
-#     print(connection_points)
-
-#     new_df["Color"] = new_df["Name"].apply(lambda name: "#39C5BB" if name in names_to_connect else "#636efa")
-
-#     fig.update_traces(marker=dict(color=new_df["Color"]))
-
-#     X = connection_points["Log Params(B)"].values.reshape(-1, 1)
-#     y = connection_points["Log Compression Rate (%)"].values
-#     model = LinearRegression().fit(X, y)
-
-#     x_min = connection_points["Log Params(B)"].min()
-#     x_max = connection_points["Log Params(B)"].max()
-#     extended_x = np.linspace(x_min, x_max * 1.5, 100)
-#     extended_x_original = np.exp(extended_x)
-#     trend_line_y = model.predict(extended_x.reshape(-1, 1))
-#     trend_line_y_original = np.exp(trend_line_y)
-
-#     trend_line = go.Scatter(
-#         x=extended_x,
-#         y=trend_line_y,
-#         mode="lines",
-#         line=dict(color="skyblue", dash="dash"),
-#         name="Trend Line",
-#         hovertemplate="<b>Params(B):</b> %{customdata[0]:.2f}<br>" + "<b>Compression Rate (%):</b> %{customdata[1]:.2f}<extra></extra>",
-#         customdata=np.stack((extended_x_original, trend_line_y_original), axis=-1),
-#     )
-
-#     fig.add_trace(trend_line)
-
-#     x_min = new_df["Params(B)"].min()
-#     x_max = new_df["Params(B)"].max()
-#     x_tick_vals = np.geomspace(x_min, x_max, num=5)
-#     x_tick_text = [f"{val:.1f}" for val in x_tick_vals]
-
-#     y_min = new_df["Compression Rate (%)"].min()
-#     y_max = new_df["Compression Rate (%)"].max()
-#     y_tick_vals = np.geomspace(y_min, y_max, num=5)
-#     y_tick_text = [f"{val:.1f}" for val in y_tick_vals]
-
-#     fig.update_xaxes(tickvals=np.log(x_tick_vals), ticktext=x_tick_text, title="Params(B)")
-#     fig.update_yaxes(tickvals=np.log(y_tick_vals), ticktext=y_tick_text, title="Compression Rate (%)", autorange="reversed")
-
-#     fig.update_layout(xaxis=dict(showgrid=True, zeroline=False), yaxis=dict(showgrid=True, zeroline=False))
-
-#     fig.update_traces(marker=dict(size=12))
-
-#     print(fig.layout)
-
-#     return fig
-
-
 def create_scaling_plot(all_data, period):
     selected_columns = ["Name", "Parameters Count (B)", "Average (The lower the better)"]
     target_data = all_data[period]
     new_df = pd.DataFrame()
-
     for size in target_data.keys():
         new_df = pd.concat([new_df, target_data[size]["cr"].loc[:, selected_columns].dropna(axis=1, how="all")], axis=0)
-
     x_values = new_df["Parameters Count (B)"].astype(float).tolist()
     y_values = new_df["Average (The lower the better)"].astype(float).tolist()
     names = new_df["Name"].tolist()
-
     x_min, x_max = np.log10(min(x_values)), np.log10(max(x_values))
     y_min, y_max = np.log10(min(y_values)), np.log10(max(y_values))
-
     x_dtick = (x_max - x_min) / 4
     y_dtick = (y_max - y_min) / 4
-
     fig = go.Figure()
-
     fig.add_trace(
         go.Scatter(
             x=x_values,
@@ -473,7 +299,6 @@ def create_scaling_plot(all_data, period):
             ),
         )
     )
-
     fig.update_layout(
         title={"text": "Compression Rate Scaling Law", "x": 0.5, "xanchor": "center", "yanchor": "top"},
         width=800,
@@ -499,10 +324,8 @@ def create_scaling_plot(all_data, period):
             autorange="reversed",
         ),
     )
-
     return fig
 
-
 def read_all_data(folder_name):
     all_data = {}
     time_list = []
@@ -517,37 +340,18 @@ def read_all_data(folder_name):
             for sheet_name in sheet_name_list:
                 final_file_name = os.path.join(folder, file_name)
                 all_data[folder_name][file_name][sheet_name] = rename_columns(pd.read_excel(final_file_name + ".xlsx", sheet_name=sheet_name))
-
     return all_data, time_list
 
-
-# def read_mutilange_data(folder_path='mutilang_data'):
-#     mutilange_data = {}
-#     excel_files = [os.path.join(folder_path, file) for file in os.listdir(folder_path) if file.endswith('.xlsx')]
-#     time_list = [file.split('.')[0] for file in excel_files]
-#     time_list = [x.split('\\')[-1] for x in time_list]
-#     for file_name in excel_files:
-#         if mutilange_data.get(file_name) is None:
-#             mutilange_data[file_name] = {}
-#         for sheet_name in sheet_name_list:
-#             mutilange_data[file_name][sheet_name] = rename_columns(
-#                 pd.read_excel(file_name, sheet_name=sheet_name))
-#     return mutilange_data, time_list
-
-
 all_data, time_list = read_all_data("data")
-# muti_lang_data, muti_lang_time_list = read_mutilange_data()
-
 time_list.sort()
 last_period = time_list[-1]
-
 initial_fig = create_scaling_plot(all_data, last_period)
 initial_metric = metric_list[0]
 initial_columns = get_unique_column_names(all_data)
 initial_colors = ["Average", "Individual Tests"]
 initial_size_range = [0, 40]
+# 初始调用 update_table 时，request 参数将为默认的 None
 initial_data = update_table(last_period, model_size_list, initial_metric, initial_columns, initial_colors, initial_size_range)
-
 css = """
 .gradio-container {
     max-width: 95% !important;
@@ -566,11 +370,11 @@ table {
     width: 100% !important;
 }
 """
-
 TITLE_HTML = '<h1 style="text-align:center"><span style="font-size:1.3em">🏆 LLM Compression Leaderboard</span></h1>'
 SUBTITLE_HTML = "<h1 style='text-align:center'><span style='font-size:0.8em'>Welcome to Uncheatable Eval LLM Compression Leaderboard, where fancy fine-tuning and cheating won't work 🚫; only compute 💻, data 📊, and real innovation 🔥 can prevail!</span></h1>"
-
-with gr.Blocks(css=css) as demo:
+# theme = gr.themes.Default(primary_hue=slate, secondary_hue=slate)
+theme = gr.themes.Default()
+with gr.Blocks(theme=theme, css=css) as demo:
     gr.HTML(TITLE_HTML)
     gr.HTML(SUBTITLE_HTML)
     with gr.Tabs() as tabs:
@@ -585,62 +389,32 @@ with gr.Blocks(css=css) as demo:
                     midpoint_slider = gr.Slider(minimum=0.1, maximum=0.9, value=0.5, step=0.01, label="Color Gradient Midpoint")
                     color_selector = gr.CheckboxGroup(label="Colored Columns", choices=["Average", "Individual Tests"], value=initial_colors)
                     colfilter = gr.CheckboxGroup(label="Data Source", choices=get_unique_column_names(all_data), value=initial_columns)
-
             table = gr.HTML(initial_data)
-
-            period_selector.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-            model_selector.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-            metric_selector.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-            colfilter.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-            color_selector.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-            size_range_slider.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-            midpoint_slider.change(
-                update_table,
-                inputs=[period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider],
-                outputs=table,
-            )
-
+            
+            # --- 核心改动点 4: 更新所有 .change() 事件，添加 gr.Request() ---
+            # 定义共享的输入列表，避免重复
+            shared_inputs = [period_selector, model_selector, metric_selector, colfilter, color_selector, size_range_slider, midpoint_slider]
+            
+            period_selector.change(update_table, inputs=shared_inputs, outputs=table)
+            model_selector.change(update_table, inputs=shared_inputs, outputs=table)
+            metric_selector.change(update_table, inputs=shared_inputs, outputs=table)
+            colfilter.change(update_table, inputs=shared_inputs, outputs=table)
+            color_selector.change(update_table, inputs=shared_inputs, outputs=table)
+            size_range_slider.change(update_table, inputs=shared_inputs, outputs=table)
+            midpoint_slider.change(update_table, inputs=shared_inputs, outputs=table)
+            
         with gr.Tab("🌍 MultiLang"):
             gr.Markdown("## Coming soon...")
-            world_languages_plot = gr.Plot(create_world_languages_gdp_chart())
-
+            # world_languages_plot = gr.Plot(create_world_languages_gdp_chart())
         with gr.Tab("📈 Scaling Law"):
             period_selector_2 = gr.Dropdown(label="Period", choices=time_list, value=last_period)
-
             def update_plot(period):
                 new_fig = create_scaling_plot(all_data, period)
                 return new_fig
-
             plot = gr.Plot(initial_fig)
             period_selector_2.change(update_plot, inputs=period_selector_2, outputs=plot)
-
         with gr.Tab("ℹ️ About"):
             gr.Markdown(read_about_md())
-
         with gr.Tab("🚀 Submit"):
             with gr.Group():
                 with gr.Row():
@@ -648,5 +422,4 @@ with gr.Blocks(css=css) as demo:
                     submit = gr.Button("Submit", variant="primary", scale=0)
             output = gr.Markdown("# Enter a public HF repo id, then hit Submit to add it to the evaluation queue.")
             submit.click(fn=submit_model, inputs=model_name, outputs=output)
-
-demo.launch(share=False)
+demo.launch(share=False)