""" Main application for Dynamic Highscores system. This file integrates all components into a unified application. """ import os import gradio as gr import threading import time from database_schema import DynamicHighscoresDB from auth import HuggingFaceAuth from benchmark_selection import BenchmarkSelector, create_benchmark_selection_ui from evaluation_queue import EvaluationQueue, create_model_submission_ui from leaderboard import Leaderboard, create_leaderboard_ui from sample_benchmarks import add_sample_benchmarks from fastapi import FastAPI, Request, HTTPException from fastapi.responses import FileResponse, HTMLResponse from fastapi.staticfiles import StaticFiles from pydantic import BaseModel # Initialize components in main thread db = DynamicHighscoresDB() auth_manager = HuggingFaceAuth(db) benchmark_selector = BenchmarkSelector(db, auth_manager) evaluation_queue = EvaluationQueue(db, auth_manager) leaderboard = Leaderboard(db) # Initialize sample benchmarks if none exist print("Checking for existing benchmarks...") benchmarks = db.get_benchmarks() if not benchmarks or len(benchmarks) == 0: print("No benchmarks found. Adding sample benchmarks...") try: # Make sure the database path is clear print(f"Database path: {db.db_path}") # Import and call the function directly num_added = add_sample_benchmarks() print(f"Added {num_added} sample benchmarks.") except Exception as e: print(f"Error adding sample benchmarks: {str(e)}") # Try direct DB insertion as fallback try: print("Attempting direct benchmark insertion...") db.add_benchmark( name="MMLU (Massive Multitask Language Understanding)", dataset_id="cais/mmlu", description="Tests knowledge across 57 subjects" ) print("Added fallback benchmark.") except Exception as inner_e: print(f"Fallback insertion failed: {str(inner_e)}") else: print(f"Found {len(benchmarks)} existing benchmarks.") # Custom CSS with theme awareness css = """ /* Theme-adaptive colored info box */ .info-text { background-color: rgba(53, 130, 220, 0.1); padding: 12px; border-radius: 8px; border-left: 4px solid #3498db; margin: 12px 0; } /* High-contrast text for elements - works in light and dark themes */ .info-text, .header, .footer, .tab-content, button, input, textarea, select, option, .gradio-container *, .markdown-text { color: var(--text-color, inherit) !important; } /* Container styling */ .container { max-width: 1200px; margin: 0 auto; } /* Header styling */ .header { text-align: center; margin-bottom: 20px; font-weight: bold; font-size: 24px; } /* Footer styling */ .footer { text-align: center; margin-top: 40px; padding: 20px; border-top: 1px solid var(--border-color-primary, #eee); } /* Login section styling */ .login-section { padding: 10px; margin-bottom: 15px; border-radius: 8px; background-color: rgba(250, 250, 250, 0.1); text-align: center; } """ # Simple manual authentication check def check_user(request: gr.Request): if request: username = request.headers.get("HF-User") if username: # User is logged in via HF Spaces print(f"User logged in: {username}") user = db.get_user_by_username(username) if not user: # Create user if they don't exist print(f"Creating new user: {username}") is_admin = (username == "Quazim0t0") db.add_user(username, username, is_admin) user = db.get_user_by_username(username) return f"Logged in as: {username}" return "Not logged in. Please login first." # Start evaluation queue worker def start_queue_worker(): # Wait a moment to ensure app is initialized time.sleep(2) try: print("Starting evaluation queue worker...") evaluation_queue.start_worker() except Exception as e: print(f"Error starting queue worker: {e}") # Create Gradio app with gr.Blocks(css=css, title="Dynamic Highscores") as app: # Display login status with gr.Row(elem_classes=["login-section"]): login_status = gr.Markdown("Checking login status...") gr.Markdown("# 🏆 Dynamic Highscores", elem_classes=["header"]) gr.Markdown(""" Welcome to Dynamic Highscores - a community benchmark platform for evaluating and comparing language models. - **Add your own benchmarks** from HuggingFace datasets - **Submit your models** for CPU-only evaluation - **Compare performance** across different models and benchmarks - **Filter results** by model type (Merge, Agent, Reasoning, Coding, etc.) """, elem_classes=["info-text"]) # Main tabs with gr.Tabs() as tabs: with gr.TabItem("📊 Leaderboard", id=0): leaderboard_ui = create_leaderboard_ui(leaderboard, db) with gr.TabItem("🚀 Submit Model", id=1): submission_ui = create_model_submission_ui(evaluation_queue, auth_manager, db) with gr.TabItem("🔍 Benchmarks", id=2): benchmark_ui = create_benchmark_selection_ui(benchmark_selector, auth_manager) gr.Markdown(""" ### About Dynamic Highscores This platform allows users to select benchmarks from HuggingFace datasets and evaluate models against them. Each user can submit one benchmark per day (admin users are exempt from this limit). All evaluations run on CPU only to ensure fair comparisons. Created by Quazim0t0 """, elem_classes=["footer"]) # Check login on page load app.load( fn=check_user, inputs=[], outputs=[login_status] ) # Create a FastAPI app to handle the static login page fastapi_app = FastAPI() # Make sure to create a static directory and put the login.html file there os.makedirs("static", exist_ok=True) # Save the login HTML to a file login_html_path = os.path.join("static", "login.html") if not os.path.exists(login_html_path): with open(login_html_path, "w") as f: f.write("""
Sign in with your HuggingFace account to submit models and benchmarks.