Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -5,29 +5,31 @@ from apscheduler.schedulers.background import BackgroundScheduler
|
|
5 |
|
6 |
# --- Make sure these imports work relative to your file structure ---
|
7 |
# Option 1: If src is a directory in the same folder as your script:
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
from src.
|
18 |
-
from src.
|
19 |
-
|
|
|
20 |
# Option 2: If you don't have these files, define placeholders (REMOVE THIS if using Option 1)
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
#
|
|
|
31 |
# --- End Placeholder Definitions ---
|
32 |
|
33 |
|
@@ -36,14 +38,14 @@ from src.submission.submit import add_new_eval # Keep if using the submit tab
|
|
36 |
# !!! IMPORTANT: Replace placeholder URLs with actual model/project pages. !!!
|
37 |
# Verify organizer and license information for accuracy.
|
38 |
data = [
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
]
|
48 |
|
49 |
# Create a master DataFrame
|
@@ -73,15 +75,17 @@ def update_leaderboard(category):
|
|
73 |
if score_column is None or score_column not in master_df.columns:
|
74 |
print(f"Warning: Invalid category '{category}' or column '{score_column}'. Falling back to default.")
|
75 |
score_column = category_to_column[DEFAULT_CATEGORY]
|
76 |
-
|
77 |
-
|
78 |
-
|
|
|
|
|
79 |
return pd.DataFrame({
|
80 |
"Rank": [],
|
81 |
"Model": [],
|
82 |
-
"
|
83 |
-
"
|
84 |
-
"
|
85 |
})
|
86 |
|
87 |
# Select base columns + the score column for sorting
|
@@ -106,14 +110,16 @@ def update_leaderboard(category):
|
|
106 |
# Rename the score column to 'Elo Score' for consistent display
|
107 |
df.rename(columns={score_column: 'Elo Score'}, inplace=True)
|
108 |
|
|
|
|
|
|
|
109 |
# Select and reorder columns for final display using the ACTUAL column names in df
|
110 |
-
# Use
|
111 |
-
final_columns = ["Rank", "Model", "
|
112 |
df = df[final_columns]
|
113 |
|
114 |
# Note: The DataFrame returned now has columns:
|
115 |
-
# 'Rank', 'Model', '
|
116 |
-
|
117 |
return df
|
118 |
|
119 |
# --- Mock/Placeholder functions/data for other tabs ---
|
@@ -125,23 +131,50 @@ pending_eval_queue_df = pd.DataFrame(columns=["Model", "Status", "Requested", "S
|
|
125 |
EVAL_COLS = ["Model", "Status", "Requested", "Started"] # Define for the dataframe headers
|
126 |
EVAL_TYPES = ["str", "str", "str", "str"] # Define for the dataframe types
|
127 |
|
128 |
-
|
129 |
# --- Keep restart function if relevant ---
|
130 |
def restart_space():
|
131 |
# Make sure REPO_ID is correctly defined/imported if this function is used
|
132 |
print(f"Attempting to restart space: {REPO_ID}")
|
133 |
# Replace with your actual space restart mechanism if needed (e.g., HfApi().restart_space(REPO_ID))
|
134 |
|
|
|
135 |
# --- Gradio App Definition ---
|
136 |
-
|
137 |
-
#
|
138 |
-
#
|
139 |
-
#
|
140 |
-
#
|
141 |
-
#
|
142 |
-
|
143 |
-
|
144 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
145 |
|
146 |
# Use a theme for better default styling
|
147 |
demo = gr.Blocks(css=custom_css, theme=gr.themes.Soft())
|
@@ -149,6 +182,7 @@ demo = gr.Blocks(css=custom_css, theme=gr.themes.Soft())
|
|
149 |
with demo:
|
150 |
# Use the TITLE variable imported or defined above
|
151 |
gr.HTML(TITLE)
|
|
|
152 |
# Use the INTRODUCTION_TEXT variable imported or defined above
|
153 |
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
|
154 |
|
@@ -165,16 +199,16 @@ with demo:
|
|
165 |
leaderboard_df_component = gr.Dataframe(
|
166 |
# Initialize with sorted data for the default category
|
167 |
value=update_leaderboard(DEFAULT_CATEGORY),
|
168 |
-
# Headers for DISPLAY
|
169 |
headers=["Rank", "Model", "Organizer", "License", "Elo Score"],
|
170 |
-
# Datatype maps to the final df columns: Rank, Model,
|
171 |
datatype=["number", "html", "str", "str", "number"],
|
172 |
interactive=False,
|
173 |
# --- FIX APPLIED: Removed unsupported 'height' argument ---
|
174 |
# row_count determines the number of rows to display
|
175 |
-
row_count=(len(master_df), "fixed"),
|
176 |
col_count=(5, "fixed"),
|
177 |
-
wrap=True, # Allow text wrapping
|
178 |
elem_id="leaderboard-table" # CSS hook for custom styling
|
179 |
)
|
180 |
# Link the radio button change to the update function
|
@@ -185,7 +219,7 @@ with demo:
|
|
185 |
)
|
186 |
|
187 |
with gr.TabItem("π About", elem_id="llm-benchmark-tab-about", id=1):
|
188 |
-
|
189 |
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
|
190 |
|
191 |
# --- Submit Tab (Commented out as in original request) ---
|
@@ -227,26 +261,45 @@ with demo:
|
|
227 |
# submission_result,
|
228 |
# )
|
229 |
|
230 |
-
|
231 |
# --- Citation Row (at the bottom, outside Tabs) ---
|
232 |
with gr.Accordion("π Citation", open=False):
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
|
|
|
|
|
|
|
|
|
|
241 |
|
242 |
# --- Keep scheduler if relevant ---
|
243 |
-
# scheduler
|
244 |
-
|
245 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
246 |
|
247 |
# --- Launch the app ---
|
248 |
# Ensures the app launches only when the script is run directly
|
249 |
if __name__ == "__main__":
|
250 |
# Ensure you have installed necessary libraries: pip install gradio pandas apscheduler
|
251 |
# Make sure your src module files (about.py etc.) are accessible OR use the placeholder definitions above.
|
|
|
252 |
demo.launch()
|
|
|
5 |
|
6 |
# --- Make sure these imports work relative to your file structure ---
|
7 |
# Option 1: If src is a directory in the same folder as your script:
|
8 |
+
try:
|
9 |
+
from src.about import (
|
10 |
+
CITATION_BUTTON_LABEL,
|
11 |
+
CITATION_BUTTON_TEXT,
|
12 |
+
EVALUATION_QUEUE_TEXT, # Keep if used by commented-out submit tab
|
13 |
+
INTRODUCTION_TEXT,
|
14 |
+
LLM_BENCHMARKS_TEXT,
|
15 |
+
TITLE,
|
16 |
+
)
|
17 |
+
from src.display.css_html_js import custom_css # Assuming this exists but might be empty
|
18 |
+
from src.envs import REPO_ID # Keep if needed for restart_space or other functions
|
19 |
+
from src.submission.submit import add_new_eval # Keep if using the submit tab
|
20 |
+
print("Successfully imported from src module.")
|
21 |
# Option 2: If you don't have these files, define placeholders (REMOVE THIS if using Option 1)
|
22 |
+
except ImportError:
|
23 |
+
print("Warning: Using placeholder values because src module imports failed.")
|
24 |
+
CITATION_BUTTON_LABEL="Citation"
|
25 |
+
CITATION_BUTTON_TEXT="Please cite us if you use this benchmark..."
|
26 |
+
EVALUATION_QUEUE_TEXT="Current evaluation queue:"
|
27 |
+
INTRODUCTION_TEXT="Welcome to the MLE-Dojo Benchmark Leaderboard."
|
28 |
+
LLM_BENCHMARKS_TEXT="Information about the benchmarks..."
|
29 |
+
TITLE="<h1>π MLE-Dojo Benchmark Leaderboard</h1>"
|
30 |
+
custom_css="" # Start with empty CSS if not imported
|
31 |
+
REPO_ID="your/space-id" # Replace with actual ID if needed
|
32 |
+
def add_new_eval(*args): return "Submission placeholder."
|
33 |
# --- End Placeholder Definitions ---
|
34 |
|
35 |
|
|
|
38 |
# !!! IMPORTANT: Replace placeholder URLs with actual model/project pages. !!!
|
39 |
# Verify organizer and license information for accuracy.
|
40 |
data = [
|
41 |
+
{'model_name': 'gpt-4o-mini', 'url': 'https://openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/', 'organizer': 'OpenAI', 'license': 'Proprietary', 'MLE-Lite_Elo': 753, 'Tabular_Elo': 839, 'NLP_Elo': 758, 'CV_Elo': 754, 'Overall': 778},
|
42 |
+
{'model_name': 'gpt-4o', 'url': 'https://openai.com/index/hello-gpt-4o/', 'organizer': 'OpenAI', 'license': 'Proprietary', 'MLE-Lite_Elo': 830, 'Tabular_Elo': 861, 'NLP_Elo': 903, 'CV_Elo': 761, 'Overall': 841},
|
43 |
+
{'model_name': 'o3-mini', 'url': 'https://openai.com/index/openai-o3-mini/', 'organizer': 'OpenAI', 'license': 'Proprietary', 'MLE-Lite_Elo': 1108, 'Tabular_Elo': 1019, 'NLP_Elo': 1056, 'CV_Elo': 1207, 'Overall': 1096}, # Fill details later
|
44 |
+
{'model_name': 'deepseek-v3', 'url': 'https://api-docs.deepseek.com/news/news1226', 'organizer': 'DeepSeek', 'license': 'DeepSeek', 'MLE-Lite_Elo': 1004, 'Tabular_Elo': 1015, 'NLP_Elo': 1028, 'CV_Elo': 1067, 'Overall': 1023},
|
45 |
+
{'model_name': 'deepseek-r1', 'url': 'https://api-docs.deepseek.com/news/news250120', 'organizer': 'DeepSeek', 'license': 'DeepSeek', 'MLE-Lite_Elo': 1137, 'Tabular_Elo': 1053, 'NLP_Elo': 1103, 'CV_Elo': 1083, 'Overall': 1100},
|
46 |
+
{'model_name': 'gemini-2.0-flash', 'url': 'https://ai.google.dev/gemini-api/docs/models#gemini-2.0-flash', 'organizer': 'Google', 'license': 'Proprietary', 'MLE-Lite_Elo': 847, 'Tabular_Elo': 923, 'NLP_Elo': 860, 'CV_Elo': 978, 'Overall': 895},
|
47 |
+
{'model_name': 'gemini-2.0-pro', 'url': 'https://blog.google/technology/google-deepmind/gemini-model-updates-february-2025/', 'organizer': 'Google', 'license': 'Proprietary', 'MLE-Lite_Elo': 1064, 'Tabular_Elo': 1139, 'NLP_Elo': 1028, 'CV_Elo': 973, 'Overall': 1054},
|
48 |
+
{'model_name': 'gemini-2.5-pro', 'url': 'https://deepmind.google/technologies/gemini/pro/', 'organizer': 'Google', 'license': 'Proprietary', 'MLE-Lite_Elo': 1257, 'Tabular_Elo': 1150, 'NLP_Elo': 1266, 'CV_Elo': 1177, 'Overall': 1214},
|
49 |
]
|
50 |
|
51 |
# Create a master DataFrame
|
|
|
75 |
if score_column is None or score_column not in master_df.columns:
|
76 |
print(f"Warning: Invalid category '{category}' or column '{score_column}'. Falling back to default.")
|
77 |
score_column = category_to_column[DEFAULT_CATEGORY]
|
78 |
+
# Check fallback column too
|
79 |
+
if score_column not in master_df.columns:
|
80 |
+
# Return empty df with correct columns if still invalid
|
81 |
+
# Use lowercase keys here consistent with master_df for the empty case
|
82 |
+
print(f"Error: Default column '{score_column}' also not found.")
|
83 |
return pd.DataFrame({
|
84 |
"Rank": [],
|
85 |
"Model": [],
|
86 |
+
"Elo Score": [],
|
87 |
+
"Organizer": [], # Changed 'organizer' -> 'Organizer' for consistency in empty case
|
88 |
+
"License": [] # Changed 'license' -> 'License' for consistency in empty case
|
89 |
})
|
90 |
|
91 |
# Select base columns + the score column for sorting
|
|
|
110 |
# Rename the score column to 'Elo Score' for consistent display
|
111 |
df.rename(columns={score_column: 'Elo Score'}, inplace=True)
|
112 |
|
113 |
+
# Rename 'organizer' and 'license' to match desired display headers
|
114 |
+
df.rename(columns={'organizer': 'Organizer', 'license': 'License'}, inplace=True)
|
115 |
+
|
116 |
# Select and reorder columns for final display using the ACTUAL column names in df
|
117 |
+
# Use capitalized 'Organizer' and 'License' here because they have been renamed.
|
118 |
+
final_columns = ["Rank", "Model", "Organizer", "License", "Elo Score"]
|
119 |
df = df[final_columns]
|
120 |
|
121 |
# Note: The DataFrame returned now has columns:
|
122 |
+
# 'Rank', 'Model', 'Organizer', 'License', 'Elo Score'
|
|
|
123 |
return df
|
124 |
|
125 |
# --- Mock/Placeholder functions/data for other tabs ---
|
|
|
131 |
EVAL_COLS = ["Model", "Status", "Requested", "Started"] # Define for the dataframe headers
|
132 |
EVAL_TYPES = ["str", "str", "str", "str"] # Define for the dataframe types
|
133 |
|
|
|
134 |
# --- Keep restart function if relevant ---
|
135 |
def restart_space():
|
136 |
# Make sure REPO_ID is correctly defined/imported if this function is used
|
137 |
print(f"Attempting to restart space: {REPO_ID}")
|
138 |
# Replace with your actual space restart mechanism if needed (e.g., HfApi().restart_space(REPO_ID))
|
139 |
|
140 |
+
|
141 |
# --- Gradio App Definition ---
|
142 |
+
|
143 |
+
# ***** FONT SIZE INCREASED HERE *****
|
144 |
+
# Add CSS rules to make the base font size larger.
|
145 |
+
# Adjust the '1.2em' value (e.g., to '1.4em', '16px') to change the size.
|
146 |
+
# The !important flag helps override theme defaults.
|
147 |
+
# If the imported custom_css already has content, append to it.
|
148 |
+
font_size_css = """
|
149 |
+
body {
|
150 |
+
font-size: 1.2em !important; /* Increase base font size */
|
151 |
+
}
|
152 |
+
/* Optional: Target specific elements if needed */
|
153 |
+
/*
|
154 |
+
#leaderboard-table th, #leaderboard-table td {
|
155 |
+
font-size: 1em !important; /* Adjust table font size relative to new body size */
|
156 |
+
padding: 10px 14px !important; /* Increase padding for better spacing */
|
157 |
+
}
|
158 |
+
h1, .markdown-text h1 { font-size: 2.2em !important; } /* Make main title larger */
|
159 |
+
h2, .markdown-text h2 { font-size: 1.8em !important; } /* Make section titles larger */
|
160 |
+
button { font-size: 1.1em !important; padding: 8px 16px !important; } /* Slightly larger buttons */
|
161 |
+
.gr-input, .gr-dropdown, .gr-textbox textarea { font-size: 1em !important; } /* Ensure inputs scale too */
|
162 |
+
*/
|
163 |
+
"""
|
164 |
+
# Append the new CSS to any existing custom_css
|
165 |
+
custom_css += font_size_css
|
166 |
+
|
167 |
+
# Add basic table styling if not already present
|
168 |
+
if "table {" not in custom_css:
|
169 |
+
custom_css += """
|
170 |
+
table { width: 100%; border-collapse: collapse; margin-top: 10px; margin-bottom: 10px; }
|
171 |
+
th, td { padding: 8px 12px; border: 1px solid #ddd; text-align: left; white-space: normal; vertical-align: top; } /* Allow wrapping, top align */
|
172 |
+
th { background-color: #f2f2f2; font-weight: bold; }
|
173 |
+
tr:nth-child(even) { background-color: #f9f9f9; }
|
174 |
+
tr:hover { background-color: #e9e9e9; }
|
175 |
+
td a { color: #007bff; text-decoration: none; }
|
176 |
+
td a:hover { text-decoration: underline; }
|
177 |
+
"""
|
178 |
|
179 |
# Use a theme for better default styling
|
180 |
demo = gr.Blocks(css=custom_css, theme=gr.themes.Soft())
|
|
|
182 |
with demo:
|
183 |
# Use the TITLE variable imported or defined above
|
184 |
gr.HTML(TITLE)
|
185 |
+
|
186 |
# Use the INTRODUCTION_TEXT variable imported or defined above
|
187 |
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
|
188 |
|
|
|
199 |
leaderboard_df_component = gr.Dataframe(
|
200 |
# Initialize with sorted data for the default category
|
201 |
value=update_leaderboard(DEFAULT_CATEGORY),
|
202 |
+
# Headers for DISPLAY should match the *renamed* columns from update_leaderboard
|
203 |
headers=["Rank", "Model", "Organizer", "License", "Elo Score"],
|
204 |
+
# Datatype maps to the final df columns: Rank, Model, Organizer, License, Elo Score
|
205 |
datatype=["number", "html", "str", "str", "number"],
|
206 |
interactive=False,
|
207 |
# --- FIX APPLIED: Removed unsupported 'height' argument ---
|
208 |
# row_count determines the number of rows to display
|
209 |
+
row_count=(len(master_df), "fixed"), # Display all rows
|
210 |
col_count=(5, "fixed"),
|
211 |
+
wrap=True, # Allow text wrapping in cells
|
212 |
elem_id="leaderboard-table" # CSS hook for custom styling
|
213 |
)
|
214 |
# Link the radio button change to the update function
|
|
|
219 |
)
|
220 |
|
221 |
with gr.TabItem("π About", elem_id="llm-benchmark-tab-about", id=1):
|
222 |
+
# Use the LLM_BENCHMARKS_TEXT variable imported or defined above
|
223 |
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
|
224 |
|
225 |
# --- Submit Tab (Commented out as in original request) ---
|
|
|
261 |
# submission_result,
|
262 |
# )
|
263 |
|
|
|
264 |
# --- Citation Row (at the bottom, outside Tabs) ---
|
265 |
with gr.Accordion("π Citation", open=False):
|
266 |
+
# Use the CITATION_BUTTON_TEXT and CITATION_BUTTON_LABEL variables imported or defined above
|
267 |
+
citation_button = gr.Textbox(
|
268 |
+
value=CITATION_BUTTON_TEXT,
|
269 |
+
label=CITATION_BUTTON_LABEL,
|
270 |
+
lines=10, # Adjust lines if needed for new font size
|
271 |
+
elem_id="citation-button",
|
272 |
+
show_copy_button=True,
|
273 |
+
)
|
274 |
+
|
275 |
+
# IGNORE_WHEN_COPYING_START
|
276 |
+
# content_copy download
|
277 |
+
# Use code with caution.
|
278 |
+
# IGNORE_WHEN_COPYING_END
|
279 |
|
280 |
# --- Keep scheduler if relevant ---
|
281 |
+
# Only start scheduler if the script is run directly
|
282 |
+
if __name__ == "__main__":
|
283 |
+
try:
|
284 |
+
scheduler = BackgroundScheduler()
|
285 |
+
# Add job only if restart_space is callable (i.e., not a placeholder or failed import)
|
286 |
+
if callable(restart_space):
|
287 |
+
# Check if REPO_ID seems valid before scheduling
|
288 |
+
if REPO_ID and REPO_ID != "your/space-id":
|
289 |
+
scheduler.add_job(restart_space, "interval", seconds=1800) # Restart every 30 mins
|
290 |
+
scheduler.start()
|
291 |
+
else:
|
292 |
+
print("Warning: REPO_ID not set or is placeholder; space restart job not scheduled.")
|
293 |
+
else:
|
294 |
+
print("Warning: restart_space function not available; space restart job not scheduled.")
|
295 |
+
except Exception as e:
|
296 |
+
print(f"Failed to initialize or start scheduler: {e}")
|
297 |
+
|
298 |
|
299 |
# --- Launch the app ---
|
300 |
# Ensures the app launches only when the script is run directly
|
301 |
if __name__ == "__main__":
|
302 |
# Ensure you have installed necessary libraries: pip install gradio pandas apscheduler
|
303 |
# Make sure your src module files (about.py etc.) are accessible OR use the placeholder definitions above.
|
304 |
+
print("Launching Gradio App...")
|
305 |
demo.launch()
|