File size: 16,708 Bytes
24a059f e109361 24a059f d24563e aeeda1a 24a059f 13e4c4d 7fcd557 c308901 13e4c4d d24563e 13e4c4d d24563e e109361 fb73da9 e109361 fb73da9 e109361 13e4c4d 6bcbc7b e109361 13e4c4d 24a059f 8b80c42 9f7748a 8b80c42 2f1a209 d67bb93 8b80c42 d67bb93 8b80c42 2f1a209 8b80c42 2f1a209 8b80c42 d30a8bb 8b80c42 d67bb93 8b80c42 2f1a209 d67bb93 9f7748a e109361 9f7748a 1fe4357 d24563e 13e4c4d d24563e 13e4c4d d24563e 8f89713 104bf5a 8f89713 1fe4357 8f89713 1fe4357 2f1a209 8f89713 e109361 1f0c8bc b8ab1fc 7bdeca8 e359f0e a1ee136 45d118c 7bdeca8 a1ee136 d55e531 a1ee136 c755c1e a1ee136 07aee39 7098daa ea94a6e d55e531 a1ee136 7098daa a1ee136 09f2be6 a1ee136 93b1517 a1ee136 cda6947 a1ee136 93b1517 07aee39 93b1517 07aee39 93b1517 a1ee136 d55e531 a1ee136 fa8abad d55e531 a1ee136 cda6947 a1ee136 cda6947 a1ee136 c755c1e a1ee136 c755c1e a1ee136 2b75ebe a1ee136 2b75ebe a1ee136 1f20775 a1ee136 1f20775 a1ee136 1f20775 a1ee136 1f20775 a1ee136 1f20775 a1ee136 1f20775 a1ee136 1f20775 79be5a0 fa8abad 5dc58a0 87904ff e359f0e 87904ff e359f0e 87904ff 3ab70ce 87904ff c912ab6 3ab70ce c912ab6 3ab70ce 87904ff e359f0e 87904ff ecfd2b0 7e020a6 ecfd2b0 e359f0e d51aeb7 ecfd2b0 f630bf2 5de4f09 f630bf2 ecfd2b0 d51aeb7 5ede32f e359f0e c2fa8d0 3be6b18 c2fa8d0 3be6b18 5ede32f 5dc58a0 5ede32f 7fcd557 3be6b18 5ede32f d24563e 73c3f03 35ad686 5dc58a0 35ad686 5dc58a0 514663d 35ad686 5dc58a0 35ad686 3d8b2b7 ecfd2b0 3ab70ce ecfd2b0 3ab70ce ecfd2b0 09f2be6 5ede32f 09f2be6 ecfd2b0 5376412 5dc58a0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 |
import gradio as gr
import pandas as pd
import os
import re
from datetime import datetime
from huggingface_hub import hf_hub_download
from huggingface_hub import HfApi, HfFolder
LEADERBOARD_FILE = "leaderboard.csv"
GROUND_TRUTH_FILE = "ground_truth.csv"
LAST_UPDATED = datetime.now().strftime("%B %d, %Y")
# Ensure authentication and suppress warnings
os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1"
HF_TOKEN = os.getenv("HF_TOKEN")
if not HF_TOKEN:
raise ValueError("HF_TOKEN environment variable is not set or invalid.")
def initialize_leaderboard_file():
"""
Ensure the leaderboard file exists and has the correct headers.
"""
if not os.path.exists(LEADERBOARD_FILE):
pd.DataFrame(columns=[
"Model Name", "Overall Accuracy", "Valid Accuracy",
"Correct Predictions", "Total Questions", "Timestamp"
]).to_csv(LEADERBOARD_FILE, index=False)
elif os.stat(LEADERBOARD_FILE).st_size == 0:
pd.DataFrame(columns=[
"Model Name", "Overall Accuracy", "Valid Accuracy",
"Correct Predictions", "Total Questions", "Timestamp"
]).to_csv(LEADERBOARD_FILE, index=False)
def clean_answer(answer):
if pd.isna(answer):
return None
answer = str(answer)
clean = re.sub(r'[^A-Da-d]', '', answer)
return clean[0].upper() if clean else None
def update_leaderboard(results):
"""
Append new submission results to the leaderboard file and push updates to the Hugging Face repository.
"""
new_entry = {
"Model Name": results['model_name'],
"Overall Accuracy": round(results['overall_accuracy'] * 100, 2),
"Valid Accuracy": round(results['valid_accuracy'] * 100, 2),
"Correct Predictions": results['correct_predictions'],
"Total Questions": results['total_questions'],
"Timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
}
try:
# Update the local leaderboard file
new_entry_df = pd.DataFrame([new_entry])
file_exists = os.path.exists(LEADERBOARD_FILE)
new_entry_df.to_csv(
LEADERBOARD_FILE,
mode='a', # Append mode
index=False,
header=not file_exists # Write header only if the file is new
)
print(f"Leaderboard updated successfully at {LEADERBOARD_FILE}")
# Push the updated file to the Hugging Face repository using HTTP API
api = HfApi()
token = HfFolder.get_token()
api.upload_file(
path_or_fileobj=LEADERBOARD_FILE,
path_in_repo="leaderboard.csv",
repo_id="SondosMB/ss", # Your Space repository
repo_type="space",
token=token
)
print("Leaderboard changes pushed to Hugging Face repository.")
except Exception as e:
print(f"Error updating leaderboard file: {e}")
def load_leaderboard():
if not os.path.exists(LEADERBOARD_FILE) or os.stat(LEADERBOARD_FILE).st_size == 0:
return pd.DataFrame({
"Model Name": [],
"Overall Accuracy": [],
"Valid Accuracy": [],
"Correct Predictions": [],
"Total Questions": [],
"Timestamp": [],
})
return pd.read_csv(LEADERBOARD_FILE)
def evaluate_predictions(prediction_file, model_name, add_to_leaderboard):
try:
ground_truth_path = hf_hub_download(
repo_id="SondosMB/ground-truth-dataset",
filename="ground_truth.csv",
repo_type="dataset",
use_auth_token=True
)
ground_truth_df = pd.read_csv(ground_truth_path)
except FileNotFoundError:
return "Ground truth file not found in the dataset repository.", load_leaderboard()
except Exception as e:
return f"Error loading ground truth: {e}", load_leaderboard()
if not prediction_file:
return "Prediction file not uploaded.", load_leaderboard()
try:
predictions_df = pd.read_csv(prediction_file.name)
merged_df = pd.merge(predictions_df, ground_truth_df, on='question_id', how='inner')
merged_df['pred_answer'] = merged_df['predicted_answer'].apply(clean_answer)
valid_predictions = merged_df.dropna(subset=['pred_answer'])
correct_predictions = (valid_predictions['pred_answer'] == valid_predictions['Answer']).sum()
total_predictions = len(merged_df)
total_valid_predictions = len(valid_predictions)
overall_accuracy = correct_predictions / total_predictions if total_predictions > 0 else 0
valid_accuracy = correct_predictions / total_valid_predictions if total_valid_predictions > 0 else 0
results = {
'model_name': model_name if model_name else "Unknown Model",
'overall_accuracy': overall_accuracy,
'valid_accuracy': valid_accuracy,
'correct_predictions': correct_predictions,
'total_questions': total_predictions,
}
if add_to_leaderboard:
update_leaderboard(results)
return "Evaluation completed and added to leaderboard.", load_leaderboard()
else:
return "Evaluation completed but not added to leaderboard.", load_leaderboard()
except Exception as e:
return f"Error during evaluation: {str(e)}", load_leaderboard()
initialize_leaderboard_file()
# Function to set default mode
# Function to set default mode
import gradio as gr
# # Custom CSS to match website style
# # Define CSS to match a modern, professional design
# # Define enhanced CSS for the entire layout
css_tech_theme = """
body {
font-family: 'Roboto', sans-serif;
background-color: #f4f6fa;
color: #333333;
margin: 0;
padding: 0;
}
/* Header Styling */
header {
text-align: center;
padding: 60px 20px;
background: linear-gradient(135deg, #6a1b9a, #64b5f6);
color: #ffffff;
border-radius: 12px;
margin-bottom: 30px;
box-shadow: 0 6px 20px rgba(0, 0, 0, 0.2);
}
header h1 {
font-size: 3.5em;
font-weight: bold;
margin-bottom: 10px;
}
header h2 {
font-size: 2em;
margin-bottom: 15px;
}
header p {
font-size: 1em;
line-height: 1.8;
}
.header-buttons {
display: flex;
justify-content: center;
gap: 15px;
margin-top: 20px;
}
.header-buttons a {
text-decoration: none;
font-size: 1.5em;
padding: 15px 30px;
border-radius: 30px;
font-weight: bold;
background: #ffffff;
color: #6a1b9a;
transition: transform 0.3s, background 0.3s;
box-shadow: 0 4px 10px rgba(0, 0, 0, 0.1);
}
.header-buttons a:hover {
background: #64b5f6;
color: #ffffff;
transform: scale(1.05);
}
/* Pre-Tabs Section */
.pre-tabs {
text-align: center;
padding: 40px 20px;
background: linear-gradient(135deg, #ffffff, #f9fafb);
border-top: 5px solid #64b5f6;
border-bottom: 5px solid #6a1b9a;
}
.pre-tabs h2, .post-tabs h2 {
font-size: 3em; /* Increase the size for better visibility */
}
.pre-tabs p, .post-tabs p {
font-size: 2.5em; /* Adjust paragraph text size */
}
.pre-tabs h2 {
color: #333333;
margin-bottom: 15px;
}
.pre-tabs p {
color: #555555;
line-height: 1.8;
}
/* Tabs Section */
.tabs {
margin: 0 auto;
padding: 20px;
background: #ffffff;
border-radius: 12px;
box-shadow: 0 4px 15px rgba(0, 0, 0, 0.1);
max-width: 1200px;
}
/* Post-Tabs Section */
.post-tabs {
text-align: center;
padding: 40px 20px;
background: linear-gradient(135deg, #64b5f6, #6a1b9a);
color: #ffffff;
border-radius: 12px;
margin-top: 30px;
}
.post-tabs h2 {
font-size: 3.4em;
margin-bottom: 15px;
}
.post-tabs p {
font-size: 2em;
line-height: 1.8;
margin-bottom: 20px;
}
.post-tabs a {
text-decoration: none;
font-size: 1.1em;
padding: 15px 30px;
border-radius: 30px;
font-weight: bold;
background: #ffffff;
color: #6a1b9a;
transition: transform 0.3s, background 0.3s;
box-shadow: 0 4px 10px rgba(0, 0, 0, 0.1);
}
.post-tabs a:hover {
background: #6a1b9a;
color: #ffffff;
transform: scale(1.05);
}
/* Footer */
footer {
background: linear-gradient(135deg, #6a1b9a, #8e44ad);
color: #ffffff;
text-align: center;
padding: 40px 20px;
margin-top: 30px;
border-radius: 12px;
box-shadow: 0 4px 10px rgba(0, 0, 0, 0.2);
}
footer h2 {
font-size: 1.5em;
margin-bottom: 15px;
}
footer p {
font-size: 1.2em;
line-height: 1.6;
margin-bottom: 20px;
}
/* Link Styling */
.social-links {
display: flex;
justify-content: center;
gap: 15px; /* Space between links */
}
.social-link {
display: inline-block;
text-decoration: none;
color: #ffffff;
background-color: #6a1b9a; /* Purple button background */
padding: 10px 20px;
border-radius: 30px;
font-size: 16px;
font-weight: bold;
transition: all 0.3s ease;
box-shadow: 0 4px 10px rgba(0, 0, 0, 0.1);
}
.social-link:hover {
background-color: #8c52d3; /* Darker shade on hover */
box-shadow: 0 6px 15px rgba(0, 0, 0, 0.2);
transform: translateY(-2px);
}
.social-link:active {
transform: translateY(1px);
box-shadow: 0 3px 8px rgba(0, 0, 0, 0.1);
}
"""
# Create the Gradio Interface
with gr.Blocks(css=css_tech_theme) as demo:
# Header Section
gr.Markdown("""
<header>
<h1>π Mobile-MMLU Challenge</h1>
<h2>π Pushing the Limits of Mobile LLMs</h2>
</header>
""")
# # Pre-Tabs Section
gr.Markdown("""
<section class="pre-tabs">
<h2>Why Participate?</h2>
<p>
The Mobile-MMLU Benchmark Competition offers a unique opportunity to evaluate your LLMs in real-world mobile scenarios. Join the challenge to drive innovation, showcase your expertise, and shape the future of mobile AI.
</p>
</section>
""")
# Tabs Section
with gr.Tabs(elem_id="tabs"):
# Overview Tab
with gr.TabItem("π Overview"):
gr.Markdown("""
<div class="tabs">
<h2>About the Competition</h2>
<p>The <strong>Mobile-MMLU Benchmark Competition</strong> is a premier challenge designed to evaluate and advance mobile-optimized Large Language Models (LLMs). It provides an unparalleled opportunity to showcase your model's ability to handle diverse, real-world scenarios while pushing the boundaries of mobile intelligence.</p>
<p>With a dataset spanning <strong>80 distinct fields</strong> and featuring <strong>16,186 questions</strong>, this competition emphasizes practical application. From education and healthcare to technology and daily life, the questions are crafted to mimic real-world challenges and test the adaptability, accuracy, and efficiency of mobile-compatible LLMs.</p>
<h3>Why Compete?</h3>
<p>Participating in this competition allows you to:
<ul>
<li>π Showcase your expertise in LLM development and optimization for mobile platforms.</li>
<li>π Benchmark your modelβs performance against others in a highly competitive environment.</li>
<li>π Contribute to advancements in AI for mobile technology, shaping the future of user-centric AI systems.</li>
</ul></p>
<h3>How It Works</h3>
<ul>
<li>1οΈβ£ <strong>Download the Dataset:</strong> Access the dataset and instructions on our
<a href="https://github.com/your-github-repo" target="_blank">GitHub page</a>.</li>
<li>2οΈβ£ <strong>Generate Predictions:</strong> Use your LLM to answer the dataset questions.
Format your predictions as a CSV file.</li>
<li>3οΈβ£ <strong>Submit Predictions:</strong> Upload your predictions on this platform.</li>
<li>4οΈβ£ <strong>Evaluation:</strong> Submissions are scored based on accuracy.</li>
<li>5οΈβ£ <strong>Leaderboard:</strong> View real-time rankings on the leaderboard.</li>
</ul>
</div>
""")
# with gr.TabItem("π€ Submission"):
# gr.Markdown("<div class='tabs'><h2>Submit Your Predictions</h2></div>")
# with gr.Row():
# file_input = gr.File(label="Upload Prediction CSV", file_types=[".csv"], interactive=True)
# model_name_input = gr.Textbox(label="Model Name", placeholder="Enter your model name")
# with gr.Row():
# overall_accuracy_display = gr.Number(label="Overall Accuracy", interactive=False)
# add_to_leaderboard_checkbox = gr.Checkbox(label="Add to Leaderboard?", value=True)
# eval_button = gr.Button("Evaluate")
# eval_status = gr.Textbox(label="Evaluation Status", interactive=False)
# def handle_evaluation(file, model_name, add_to_leaderboard):
# status, leaderboard = evaluate_predictions(file, model_name, add_to_leaderboard)
# if leaderboard.empty:
# overall_accuracy = 0
# else:
# overall_accuracy = leaderboard.iloc[-1]["Overall Accuracy"]
# return status, overall_accuracy
# eval_button.click(
# handle_evaluation,
# inputs=[file_input, model_name_input, add_to_leaderboard_checkbox],
# outputs=[eval_status, overall_accuracy_display],
# )
with gr.TabItem("π€ Submission"):
with gr.Row():
file_input = gr.File(label="Upload Prediction CSV", file_types=[".csv"], interactive=True)
model_name_input = gr.Textbox(label="Model Name", placeholder="Enter your model name")
with gr.Row():
overall_accuracy_display = gr.Number(label="Overall Accuracy", interactive=False)
eval_button = gr.Button("Evaluate")
eval_status = gr.Textbox(label="Evaluation Status", interactive=False)
submit_button = gr.Button("Prove and Submit to Leaderboard", visible=False) # Initially hidden
def handle_evaluation(file, model_name):
# Perform evaluation
status, leaderboard = evaluate_predictions(file, model_name, add_to_leaderboard=False)
if leaderboard.empty:
overall_accuracy = 0
else:
overall_accuracy = leaderboard.iloc[-1]["Overall Accuracy"]
# Show the submit button after evaluation
return status, overall_accuracy, gr.update(visible=True)
def handle_submission(file, model_name):
# Handle leaderboard submission
status, _ = evaluate_predictions(file, model_name, add_to_leaderboard=True)
return f"Submission to leaderboard completed: {status}"
eval_button.click(
handle_evaluation,
inputs=[file_input, model_name_input],
outputs=[eval_status, overall_accuracy_display, submit_button],)
submit_button.click(
handle_submission,
inputs=[file_input, model_name_input],
outputs=[eval_status],)
with gr.TabItem("π
Leaderboard"):
leaderboard_table = gr.Dataframe(
value=load_leaderboard(),
label="Leaderboard",
interactive=False,
wrap=True,
)
refresh_button = gr.Button("Refresh Leaderboard")
refresh_button.click(
lambda: load_leaderboard(),
inputs=[],
outputs=[leaderboard_table],
)
# Post-Tabs Section
gr.Markdown("""
<section class="post-tabs">
<h2>Ready to Compete?</h2>
<h3>
Submit your predictions today and make your mark in advancing mobile AI technologies.
Show the world what your model can achieve!
<h3>
</section>
""")
# Footer Section
gr.Markdown("""
<footer>
<h2>Stay Connected</h2>
<p>
Follow us on social media or contact us for any queries. Let's shape the future of AI together!
</p>
<div class="social-links">
<a href="https://website.com" target="_blank" class="social-link">π Website</a>
<a href="https://github.com/VILA-Lab/Mobile-MMLU" target="_blank" class="social-link">π GitHub</a>
</div>
</footer>
""")
demo.launch()
|