update

app.py

@@ -297,25 +297,50 @@ def save_leaderboard(df: pd.DataFrame):
     # Upload to private HuggingFace repo
     try:
         token = os.environ.get('HF_TOKEN')
-        if token:
-            print("⏳ Uploading leaderboard to private repository...")
+        if not token:
+            print("⚠️  HF_TOKEN not found in environment")
+            print("   Set HF_TOKEN secret in Space settings to enable repo sync")
+            print("   Leaderboard saved locally only (will not persist across restarts)")
+            return
 
-            api = HfApi()
-            api.upload_file(
-                path_or_fileobj=str(LEADERBOARD_FILE),
-                path_in_repo="leaderboard.json",
-                repo_id="UIIAmerica/MedVidBench-GroundTruth",
-                repo_type="dataset",
-                token=token,
-                commit_message=f"Update leaderboard: {len(df)} entries ({datetime.now().strftime('%Y-%m-%d %H:%M:%S')})"
-            )
+        print("⏳ Uploading leaderboard to private repository...")
+        print(f"   Target: UIIAmerica/MedVidBench-GroundTruth/leaderboard.json")
+        print(f"   Entries: {len(df)}")
+
+        api = HfApi()
+
+        # Upload with detailed error handling
+        result = api.upload_file(
+            path_or_fileobj=str(LEADERBOARD_FILE),
+            path_in_repo="leaderboard.json",
+            repo_id="UIIAmerica/MedVidBench-GroundTruth",
+            repo_type="dataset",
+            token=token,
+            commit_message=f"Update leaderboard: {len(df)} entries ({datetime.now().strftime('%Y-%m-%d %H:%M:%S')})"
+        )
+
+        print(f"✓ Successfully uploaded leaderboard to private repo")
+        print(f"   Commit URL: {result}")
 
-            print(f"✓ Uploaded leaderboard to private repo")
-        else:
-            print("⚠️  HF_TOKEN not found, skipping upload to private repo")
     except Exception as e:
-        print(f"⚠️  Failed to upload leaderboard to private repo: {e}")
-        print("   Leaderboard is still saved locally")
+        error_msg = str(e)
+        print(f"❌ Failed to upload leaderboard to private repo")
+        print(f"   Error: {error_msg}")
+
+        # Provide specific guidance based on error type
+        if "401" in error_msg or "Unauthorized" in error_msg:
+            print("   → Issue: Invalid or expired token")
+            print("   → Fix: Regenerate HF_TOKEN with write permission")
+        elif "404" in error_msg or "Not Found" in error_msg:
+            print("   → Issue: Repository not found")
+            print("   → Fix: Create UIIAmerica/MedVidBench-GroundTruth repo")
+        elif "403" in error_msg or "Forbidden" in error_msg:
+            print("   → Issue: Token lacks write permission")
+            print("   → Fix: Use token with write access to dataset")
+        else:
+            print(f"   → Check HuggingFace status and repo permissions")
+
+        print("   ⚠️  Leaderboard saved locally only (will not persist)")
 
 
 def backup_results_to_repo(model_name: str, results_dir: Path):

test_hf_token.py

@@ -0,0 +1,157 @@
+"""Test HF_TOKEN configuration for repository access.
+
+Run this script to verify that HF_TOKEN is properly configured
+and has the necessary permissions for the private repo.
+"""
+
+import os
+import sys
+from huggingface_hub import HfApi
+from pathlib import Path
+
+REPO_ID = "UIIAmerica/MedVidBench-GroundTruth"
+REPO_TYPE = "dataset"
+
+def test_hf_token():
+    """Test HF_TOKEN configuration."""
+    print("=" * 80)
+    print("TESTING HF_TOKEN CONFIGURATION")
+    print("=" * 80)
+
+    # Check if token exists
+    print("\n[1/4] Checking HF_TOKEN environment variable...")
+    token = os.environ.get('HF_TOKEN')
+
+    if not token:
+        print("❌ FAILED: HF_TOKEN not found in environment")
+        print("\nHow to fix:")
+        print("1. Generate token at: https://huggingface.co/settings/tokens")
+        print("2. Grant 'write' permission to repositories")
+        print("3. Set as environment variable:")
+        print("   export HF_TOKEN='your_token_here'")
+        print("\n4. For HuggingFace Spaces:")
+        print("   Settings → Repository secrets → Add HF_TOKEN")
+        sys.exit(1)
+
+    print(f"✓ HF_TOKEN found (length: {len(token)} chars)")
+    masked_token = token[:7] + "..." + token[-4:] if len(token) > 11 else "***"
+    print(f"  Token: {masked_token}")
+
+    # Initialize API
+    print("\n[2/4] Initializing HuggingFace API...")
+    try:
+        api = HfApi()
+        print("✓ HfApi initialized")
+    except Exception as e:
+        print(f"❌ FAILED: {e}")
+        sys.exit(1)
+
+    # Test repository access (read)
+    print(f"\n[3/4] Testing READ access to {REPO_ID}...")
+    try:
+        repo_info = api.repo_info(
+            repo_id=REPO_ID,
+            repo_type=REPO_TYPE,
+            token=token
+        )
+        print(f"✓ Successfully accessed repository")
+        print(f"  Repo: {repo_info.id}")
+        print(f"  Private: {repo_info.private}")
+        print(f"  Last modified: {repo_info.lastModified}")
+
+        # List files
+        files = api.list_repo_files(
+            repo_id=REPO_ID,
+            repo_type=REPO_TYPE,
+            token=token
+        )
+        print(f"  Files in repo: {len(files)}")
+        for file in files:
+            print(f"    - {file}")
+
+    except Exception as e:
+        error_msg = str(e)
+        print(f"❌ FAILED: {error_msg}")
+
+        if "401" in error_msg or "Unauthorized" in error_msg:
+            print("\n→ Issue: Invalid or expired token")
+            print("→ Fix: Regenerate token at https://huggingface.co/settings/tokens")
+        elif "404" in error_msg or "Not Found" in error_msg:
+            print(f"\n→ Issue: Repository '{REPO_ID}' not found")
+            print("→ Fix: Create the repository:")
+            print(f"  1. Go to: https://huggingface.co/new-dataset")
+            print(f"  2. Owner: UIIAmerica")
+            print(f"  3. Name: MedVidBench-GroundTruth")
+            print(f"  4. Visibility: Private")
+        elif "403" in error_msg or "Forbidden" in error_msg:
+            print("\n→ Issue: No access to private repository")
+            print("→ Fix: Ensure you're a member of UIIAmerica organization")
+
+        sys.exit(1)
+
+    # Test write access
+    print(f"\n[4/4] Testing WRITE access to {REPO_ID}...")
+    try:
+        # Create a test file
+        test_file = Path("test_upload.txt")
+        with open(test_file, 'w') as f:
+            f.write("Test upload to verify write permissions\n")
+
+        print("  Creating test file...")
+        result = api.upload_file(
+            path_or_fileobj=str(test_file),
+            path_in_repo="test_upload.txt",
+            repo_id=REPO_ID,
+            repo_type=REPO_TYPE,
+            token=token,
+            commit_message="Test write access"
+        )
+
+        print(f"✓ Successfully uploaded test file")
+        print(f"  Commit: {result}")
+
+        # Clean up test file
+        print("  Cleaning up test file...")
+        api.delete_file(
+            path_in_repo="test_upload.txt",
+            repo_id=REPO_ID,
+            repo_type=REPO_TYPE,
+            token=token,
+            commit_message="Remove test file"
+        )
+        test_file.unlink()
+
+        print(f"✓ Successfully deleted test file")
+
+    except Exception as e:
+        error_msg = str(e)
+        print(f"❌ FAILED: {error_msg}")
+
+        if "403" in error_msg or "Forbidden" in error_msg:
+            print("\n→ Issue: Token does not have write permission")
+            print("→ Fix:")
+            print("  1. Go to: https://huggingface.co/settings/tokens")
+            print("  2. Create new token with WRITE permission")
+            print("  3. Update HF_TOKEN environment variable")
+        elif "401" in error_msg:
+            print("\n→ Issue: Token invalid for write operations")
+            print("→ Fix: Ensure token has 'write' scope")
+
+        sys.exit(1)
+
+    # Success
+    print("\n" + "=" * 80)
+    print("✅ ALL TESTS PASSED")
+    print("=" * 80)
+    print("\nYour HF_TOKEN is correctly configured with:")
+    print("  ✓ Valid authentication")
+    print("  ✓ Read access to private repository")
+    print("  ✓ Write access to private repository")
+    print("\nYou can now:")
+    print("  1. Deploy app.py to HuggingFace Spaces")
+    print("  2. Add HF_TOKEN as a Space secret")
+    print("  3. Leaderboard will automatically sync to private repo")
+
+
+if __name__ == "__main__":
+    test_hf_token()

upload_initial_data.py

@@ -0,0 +1,128 @@
+"""Upload initial data to private HuggingFace repository.
+
+This script uploads:
+1. ground_truth.json - Private test set
+2. leaderboard.json - Initial leaderboard with MedGRPO model
+
+Run this once during initial setup.
+"""
+
+import os
+import sys
+import json
+from pathlib import Path
+from huggingface_hub import HfApi
+
+# Configuration
+REPO_ID = "UIIAmerica/MedVidBench-GroundTruth"
+REPO_TYPE = "dataset"
+
+def create_initial_leaderboard():
+    """Create initial leaderboard.json with MedGRPO model data."""
+    leaderboard_data = [
+        {
+            "rank": 1,
+            "model_name": "Qwen2.5-VL-7B-MedGRPO",
+            "organization": "UII",
+            "cvs_acc": 0.914,
+            "nap_acc": 0.427,
+            "sa_acc": 0.244,
+            "stg_miou": 0.202,
+            "tag_miou_03": 0.216,
+            "tag_miou_05": 0.156,
+            "dvc_llm": 3.797,
+            "dvc_f1": 0.210,
+            "vs_llm": 4.184,
+            "rc_llm": 3.442,
+            "date": "2025-01-14",
+            "contact": "gaozhongpai@gmail.com"
+        }
+    ]
+
+    leaderboard_file = Path("leaderboard.json")
+    with open(leaderboard_file, 'w') as f:
+        json.dump(leaderboard_data, f, indent=2)
+
+    print(f"✓ Created leaderboard.json with 1 entry (Qwen2.5-VL-7B-MedGRPO)")
+    return leaderboard_file
+
+
+def main():
+    """Upload initial files to private repo."""
+    # Check token
+    token = os.environ.get('HF_TOKEN')
+    if not token:
+        print("❌ HF_TOKEN environment variable not set")
+        print("   Please run: export HF_TOKEN='your_token_here'")
+        sys.exit(1)
+
+    print("=" * 80)
+    print(f"UPLOADING INITIAL DATA TO {REPO_ID}")
+    print("=" * 80)
+
+    api = HfApi()
+
+    # 1. Upload ground truth
+    print("\n[1/2] Uploading ground_truth.json...")
+    ground_truth_file = Path("data/ground_truth.json")
+
+    if not ground_truth_file.exists():
+        print(f"   ❌ File not found: {ground_truth_file}")
+        print(f"   Skipping ground_truth.json upload...")
+    else:
+        file_size = ground_truth_file.stat().st_size / (1024 * 1024)  # MB
+        print(f"   File size: {file_size:.2f} MB")
+
+        try:
+            api.upload_file(
+                path_or_fileobj=str(ground_truth_file),
+                path_in_repo="ground_truth.json",
+                repo_id=REPO_ID,
+                repo_type=REPO_TYPE,
+                token=token,
+                commit_message="Upload ground truth data"
+            )
+            print(f"   ✓ Uploaded ground_truth.json")
+        except Exception as e:
+            print(f"   ❌ Failed: {e}")
+            sys.exit(1)
+
+    # 2. Upload leaderboard
+    print("\n[2/2] Uploading leaderboard.json...")
+
+    # Create leaderboard with MedGRPO data
+    leaderboard_file = create_initial_leaderboard()
+
+    file_size = leaderboard_file.stat().st_size
+    print(f"   File size: {file_size} bytes")
+
+    try:
+        api.upload_file(
+            path_or_fileobj=str(leaderboard_file),
+            path_in_repo="leaderboard.json",
+            repo_id=REPO_ID,
+            repo_type=REPO_TYPE,
+            token=token,
+            commit_message="Initialize leaderboard with Qwen2.5-VL-7B-MedGRPO"
+        )
+        print(f"   ✓ Uploaded leaderboard.json")
+    except Exception as e:
+        print(f"   ❌ Failed: {e}")
+        sys.exit(1)
+
+    print("\n" + "=" * 80)
+    print("✅ UPLOAD COMPLETE")
+    print("=" * 80)
+    print(f"\nRepository: https://huggingface.co/datasets/{REPO_ID}")
+    print("\nUploaded:")
+    print("  ✓ ground_truth.json (if available)")
+    print("  ✓ leaderboard.json with Qwen2.5-VL-7B-MedGRPO")
+    print("\nNext steps:")
+    print("1. Verify files in repository")
+    print("2. Add HF_TOKEN secret to HuggingFace Space")
+    print("3. Deploy app.py to Space")
+    print("4. Check app logs for successful loading")
+
+
+if __name__ == "__main__":
+    main()