File size: 15,895 Bytes
7fcd557 d24563e 6bcbc7b d24563e 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 7fcd557 d51aeb7 a45bd57 d51aeb7 d24563e 7fcd557 d51aeb7 1f0c8bc c2fa8d0 54aa8d9 c2fa8d0 d51aeb7 a45bd57 c2fa8d0 a45bd57 c2fa8d0 1f0c8bc c2fa8d0 7fcd557 c2fa8d0 fca838b 7fcd557 fca838b 1fe4357 c2fa8d0 7fcd557 d24563e 7fcd557 8f89713 7fcd557 d24563e |
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 |
# # demo.launch()
# import gradio as gr
# import pandas as pd
# import os
# import re
# from datetime import datetime
# LEADERBOARD_FILE = "leaderboard.csv" # File to store all submissions persistently
# LAST_UPDATED = datetime.now().strftime("%B %d, %Y")
# def initialize_leaderboard_file():
# """
# Ensure the leaderboard file exists and has the correct headers.
# """
# if not os.path.exists(LEADERBOARD_FILE):
# # Create the file with headers
# pd.DataFrame(columns=[
# "Model Name", "Overall Accuracy", "Valid Accuracy",
# "Correct Predictions", "Total Questions", "Timestamp"
# ]).to_csv(LEADERBOARD_FILE, index=False)
# else:
# # Check if the file is empty and write headers if needed
# if 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):
# """
# Clean and normalize the predicted answers.
# """
# if pd.isna(answer):
# return None
# answer = str(answer)
# clean = re.sub(r'[^A-Da-d]', '', answer)
# if clean:
# return clean[0].upper()
# return None
# def update_leaderboard(results):
# """
# Append new submission results to the leaderboard file.
# """
# 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"),
# }
# new_entry_df = pd.DataFrame([new_entry])
# new_entry_df.to_csv(LEADERBOARD_FILE, mode='a', index=False, header=False)
# def load_leaderboard():
# """
# Load all submissions from the leaderboard file.
# """
# 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):
# """
# Evaluate predictions and optionally add results to the leaderboard.
# """
# ground_truth_file = "ground_truth.csv"
# if not os.path.exists(ground_truth_file):
# return "Ground truth file not found.", load_leaderboard()
# if not prediction_file:
# return "Prediction file not uploaded.", load_leaderboard()
# try:
# # Load predictions and ground truth
# predictions_df = pd.read_csv(prediction_file.name)
# ground_truth_df = pd.read_csv(ground_truth_file)
# # Merge predictions with ground truth
# 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)
# # Evaluate predictions
# 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)
# # Calculate accuracy
# 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,
# }
# # Update leaderboard only if opted in
# 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
# initialize_leaderboard_file()
# # Gradio Interface
# with gr.Blocks() as demo:
# gr.Markdown("# Prediction Evaluation Tool with Leaderboard")
# with gr.Tabs():
# # Submission Tab
# with gr.TabItem("🏅 Submission"):
# file_input = gr.File(label="Upload Prediction CSV")
# model_name_input = gr.Textbox(label="Model Name", placeholder="Enter your model name")
# add_to_leaderboard_checkbox = gr.Checkbox(label="Add to Leaderboard?", value=True)
# eval_status = gr.Textbox(label="Evaluation Status", interactive=False)
# leaderboard_table_preview = gr.Dataframe(
# value=load_leaderboard(),
# label="Leaderboard (Preview)",
# interactive=False,
# wrap=True,
# )
# eval_button = gr.Button("Evaluate and Update Leaderboard")
# eval_button.click(
# evaluate_predictions,
# inputs=[file_input, model_name_input, add_to_leaderboard_checkbox],
# outputs=[eval_status, leaderboard_table_preview],
# )
# # Leaderboard Tab
# with gr.TabItem("🏅 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],
# )
# gr.Markdown(f"Last updated on **{LAST_UPDATED}**")
# demo.launch()
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()
with gr.Blocks() as demo:
gr.Markdown("""
# Competition Title
### Welcome to the Competition Overview
)
Here you can submit your predictions, view the leaderboard, and track your performance!
""")
with gr.Tabs():
with gr.TabItem("📖 Overview"):
gr.Markdown("""
## Overview
# Welcome to the Mobile-MMLU Benchmark Competition
Evaluate the performance of mobile-compatible Large Language Models (LLMs) on 16,186 scenario-based and factual questions across 80 fields. Compete to showcase your model’s accuracy for real-world mobile scenarios.
## What is Mobile-MMLU?
Mobile-MMLU is a benchmark designed to test the capabilities of LLMs optimized for mobile use. By participating in this competition, you contribute to advancing mobile intelligence benchmarks and shaping the future of mobile-compatible AI systems.
---
## How It Works
1. **Download the Dataset**
Access the dataset and detailed generation instructions on our [GitHub page](https://github.com/your-github-repo).
2. **Generate Predictions**
Use your LLM to answer the questions and format your predictions as a CSV file with the following structure as written on our GitHub page :
3. **Submit Predictions**
Upload your predictions via the submission portal.
4. **Evaluation**
Your submission will be scored on accuracy
5. **Leaderboard**
Compare your results against other participants on the live leaderboard.
---
## Competition Tasks
Participants are tasked with generating predictions for the dataset and optimizing their models for:
- **Accuracy**: Correctly answering questions across diverse fields.
---
## Get Started
1. **Prepare Your Model**
Refer to our [GitHub page](https://github.com/your-github-repo) for dataset access and response generation instructions.
2. **Submit Predictions**
Format your submission as specified in the rules.
3. **Track Progress**
Check the leaderboard for real-time rankings.
---
## Contact Us
For questions or support, contact us at: [Insert Email Address]
""")
with gr.TabItem("📤 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)
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("🏅 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],
)
gr.Markdown(f"Last updated on **{LAST_UPDATED}**")
demo.launch()
|