Update app.py

app.py

@@ -1118,6 +1118,10 @@ Apache 2.0 (inherited from original model)
 # 업로드 전 검증 함수 (신규!)
 # =====================================================
 
+# =====================================================
+# 업로드 전 검증 함수 (수정!)
+# =====================================================
+
 def verify_phoenix_model_before_upload(model_path: str) -> Tuple[bool, str, Dict]:
     """
     Upload 전 PHOENIX 모델 검증
@@ -1128,13 +1132,33 @@ def verify_phoenix_model_before_upload(model_path: str) -> Tuple[bool, str, Dict
     print("\n🧪 Pre-upload Verification...")
     
     try:
-        # 1. 파일 존재 확인
+        # 1. 파일 존재 확인 (safetensors OR pytorch_model.bin)
         model_path = Path(model_path)
-        required_files = ['config.json', 'modeling_phoenix.py', 'pytorch_model.bin', 'README.md']
         
-        for file in required_files:
-            if not (model_path / file).exists():
-                return False, f"❌ Missing file: {file}", {}
+        # 필수 파일 체크 (모델 가중치는 둘 중 하나만 있으면 됨)
+        config_exists = (model_path / 'config.json').exists()
+        modeling_exists = (model_path / 'modeling_phoenix.py').exists()
+        readme_exists = (model_path / 'README.md').exists()
+        
+        # 모델 가중치 파일 확인 (safetensors 우선)
+        safetensors_exists = (model_path / 'model.safetensors').exists()
+        pytorch_bin_exists = (model_path / 'pytorch_model.bin').exists()
+        model_weights_exist = safetensors_exists or pytorch_bin_exists
+        
+        print(f"   📄 File Check:")
+        print(f"      config.json: {'✅' if config_exists else '❌'}")
+        print(f"      modeling_phoenix.py: {'✅' if modeling_exists else '❌'}")
+        print(f"      README.md: {'✅' if readme_exists else '❌'}")
+        print(f"      model weights: {'✅ (safetensors)' if safetensors_exists else '✅ (pytorch_model.bin)' if pytorch_bin_exists else '❌'}")
+        
+        if not config_exists:
+            return False, "❌ Missing file: config.json", {}
+        if not modeling_exists:
+            return False, "❌ Missing file: modeling_phoenix.py", {}
+        if not readme_exists:
+            return False, "❌ Missing file: README.md", {}
+        if not model_weights_exist:
+            return False, "❌ Missing model weights (need model.safetensors or pytorch_model.bin)", {}
         
         print("   ✅ All required files present")
         
@@ -1251,6 +1275,7 @@ def verify_phoenix_model_before_upload(model_path: str) -> Tuple[bool, str, Dict
             'total_layers': total_layers,
             'retention_rate': retention_rate,
             'generation_quality': avg_score,
+            'model_format': 'safetensors' if safetensors_exists else 'pytorch_bin'
         }
         
         print("\n✅ Pre-upload verification PASSED!")
@@ -1264,137 +1289,7 @@ def verify_phoenix_model_before_upload(model_path: str) -> Tuple[bool, str, Dict
 
 
 # =====================================================
-# 데이터베이스
-# =====================================================
-
-class ExperimentDatabase:
-    """SQLite database with migration support"""
-    
-    def __init__(self, db_path: str):
-        self.db_path = db_path
-        self.init_database()
-        self.migrate_database()
-    
-    def init_database(self):
-        with sqlite3.connect(self.db_path) as conn:
-            cursor = conn.cursor()
-            cursor.execute("""
-                CREATE TABLE IF NOT EXISTS experiments (
-                    id INTEGER PRIMARY KEY AUTOINCREMENT,
-                    model_type TEXT NOT NULL,
-                    sequence_length INTEGER,
-                    use_hierarchical BOOLEAN,
-                    attention_replaced BOOLEAN,
-                    layers_converted INTEGER,
-                    total_layers INTEGER,
-                    elapsed_time REAL,
-                    memory_mb REAL,
-                    throughput REAL,
-                    config_json TEXT,
-                    metrics_json TEXT,
-                    timestamp DATETIME DEFAULT CURRENT_TIMESTAMP
-                )
-            """)
-            
-            cursor.execute("""
-                CREATE TABLE IF NOT EXISTS burning_history (
-                    id INTEGER PRIMARY KEY AUTOINCREMENT,
-                    model_url TEXT NOT NULL,
-                    output_path TEXT NOT NULL,
-                    hub_url TEXT,
-                    use_hierarchical BOOLEAN,
-                    dataset_used BOOLEAN,
-                    conversion_rate REAL,
-                    training_steps INTEGER,
-                    final_loss REAL,
-                    evaluation_score REAL,
-                    verification_passed BOOLEAN,
-                    timestamp DATETIME DEFAULT CURRENT_TIMESTAMP
-                )
-            """)
-            conn.commit()
-    
-    def migrate_database(self):
-        with sqlite3.connect(self.db_path) as conn:
-            cursor = conn.cursor()
-            cursor.execute("PRAGMA table_info(burning_history)")
-            columns = [col[1] for col in cursor.fetchall()]
-            
-            if 'hub_url' not in columns:
-                print("🔄 Migrating database: Adding hub_url column...")
-                cursor.execute("ALTER TABLE burning_history ADD COLUMN hub_url TEXT")
-            
-            if 'verification_passed' not in columns:
-                print("🔄 Migrating database: Adding verification_passed column...")
-                cursor.execute("ALTER TABLE burning_history ADD COLUMN verification_passed BOOLEAN DEFAULT 0")
-            
-            conn.commit()
-            print("✅ Database migration complete!")
-    
-    def save_experiment(self, config: Dict, metrics: Dict) -> int:
-        with sqlite3.connect(self.db_path) as conn:
-            cursor = conn.cursor()
-            cursor.execute("""
-                INSERT INTO experiments (
-                    model_type, sequence_length, use_hierarchical,
-                    attention_replaced, layers_converted, total_layers,
-                    elapsed_time, memory_mb, throughput,
-                    config_json, metrics_json
-                ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
-            """, (
-                config.get('model_type'),
-                config.get('sequence_length'),
-                config.get('use_hierarchical'),
-                config.get('attention_replaced'),
-                config.get('layers_converted'),
-                config.get('total_layers'),
-                metrics.get('elapsed_time'),
-                metrics.get('memory_mb'),
-                metrics.get('throughput'),
-                json.dumps(config),
-                json.dumps(metrics)
-            ))
-            conn.commit()
-            return cursor.lastrowid
-    
-    def save_burning(self, burning_info: Dict) -> int:
-        with sqlite3.connect(self.db_path) as conn:
-            cursor = conn.cursor()
-            cursor.execute("""
-                INSERT INTO burning_history (
-                    model_url, output_path, hub_url, use_hierarchical,
-                    dataset_used, conversion_rate, training_steps,
-                    final_loss, evaluation_score, verification_passed
-                ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
-            """, (
-                burning_info.get('model_url'),
-                burning_info.get('output_path'),
-                burning_info.get('hub_url'),
-                burning_info.get('use_hierarchical'),
-                burning_info.get('dataset_used'),
-                burning_info.get('conversion_rate'),
-                burning_info.get('training_steps', 0),
-                burning_info.get('final_loss'),
-                burning_info.get('evaluation_score'),
-                burning_info.get('verification_passed', False),
-            ))
-            conn.commit()
-            return cursor.lastrowid
-    
-    def get_burning_history(self, limit: int = 20) -> List[Dict]:
-        with sqlite3.connect(self.db_path) as conn:
-            conn.row_factory = sqlite3.Row
-            cursor = conn.cursor()
-            cursor.execute("SELECT * FROM burning_history ORDER BY timestamp DESC LIMIT ?", (limit,))
-            return [dict(row) for row in cursor.fetchall()]
-
-
-# =====================================================
-# HuggingFace Hub Upload (검증 통합!)
-# =====================================================
-
-# =====================================================
-# HuggingFace Hub Upload (개선!)
+# HuggingFace Hub Upload (수정!)
 # =====================================================
 
 def upload_to_huggingface_hub(
@@ -1445,6 +1340,7 @@ def upload_to_huggingface_hub(
         print(f"✅ Pre-upload verification PASSED!")
         print(f"   Retention Rate: {metrics.get('retention_rate', 0)*100:.1f}%")
         print(f"   Generation Quality: {metrics.get('generation_quality', 0):.2f}/1.00")
+        print(f"   Model Format: {metrics.get('model_format', 'unknown')}")
     else:
         print("\n⚠️ Skipping pre-upload verification")
     
@@ -1493,13 +1389,18 @@ def upload_to_huggingface_hub(
         print(f"\n📤 Uploading files to HuggingFace Hub...")
         print(f"   This may take a few minutes depending on model size...")
         
-        required_files = ['config.json', 'modeling_phoenix.py', 'pytorch_model.bin']
-        for file in required_files:
-            file_path = model_path / file
-            if not file_path.exists():
-                error_msg = f"❌ Required file missing: {file}"
-                print(f"\n{error_msg}")
-                return False, "", error_msg
+        # 필수 파일 체크 (safetensors OR pytorch_model.bin)
+        config_exists = (model_path / 'config.json').exists()
+        modeling_exists = (model_path / 'modeling_phoenix.py').exists()
+        safetensors_exists = (model_path / 'model.safetensors').exists()
+        pytorch_bin_exists = (model_path / 'pytorch_model.bin').exists()
+        
+        if not config_exists:
+            return False, "", "❌ config.json not found"
+        if not modeling_exists:
+            return False, "", "❌ modeling_phoenix.py not found"
+        if not (safetensors_exists or pytorch_bin_exists):
+            return False, "", "❌ Model weights not found (need model.safetensors or pytorch_model.bin)"
         
         print(f"✅ All required files present")
         
@@ -1539,6 +1440,135 @@ def upload_to_huggingface_hub(
         return False, "", f"❌ Upload failed: {str(e)}\n\nFull error:\n{error_msg}"
 
 
+# =====================================================
+# 데이터베이스
+# =====================================================
+
+class ExperimentDatabase:
+    """SQLite database with migration support"""
+    
+    def __init__(self, db_path: str):
+        self.db_path = db_path
+        self.init_database()
+        self.migrate_database()
+    
+    def init_database(self):
+        with sqlite3.connect(self.db_path) as conn:
+            cursor = conn.cursor()
+            cursor.execute("""
+                CREATE TABLE IF NOT EXISTS experiments (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    model_type TEXT NOT NULL,
+                    sequence_length INTEGER,
+                    use_hierarchical BOOLEAN,
+                    attention_replaced BOOLEAN,
+                    layers_converted INTEGER,
+                    total_layers INTEGER,
+                    elapsed_time REAL,
+                    memory_mb REAL,
+                    throughput REAL,
+                    config_json TEXT,
+                    metrics_json TEXT,
+                    timestamp DATETIME DEFAULT CURRENT_TIMESTAMP
+                )
+            """)
+            
+            cursor.execute("""
+                CREATE TABLE IF NOT EXISTS burning_history (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    model_url TEXT NOT NULL,
+                    output_path TEXT NOT NULL,
+                    hub_url TEXT,
+                    use_hierarchical BOOLEAN,
+                    dataset_used BOOLEAN,
+                    conversion_rate REAL,
+                    training_steps INTEGER,
+                    final_loss REAL,
+                    evaluation_score REAL,
+                    verification_passed BOOLEAN,
+                    timestamp DATETIME DEFAULT CURRENT_TIMESTAMP
+                )
+            """)
+            conn.commit()
+    
+    def migrate_database(self):
+        with sqlite3.connect(self.db_path) as conn:
+            cursor = conn.cursor()
+            cursor.execute("PRAGMA table_info(burning_history)")
+            columns = [col[1] for col in cursor.fetchall()]
+            
+            if 'hub_url' not in columns:
+                print("🔄 Migrating database: Adding hub_url column...")
+                cursor.execute("ALTER TABLE burning_history ADD COLUMN hub_url TEXT")
+            
+            if 'verification_passed' not in columns:
+                print("🔄 Migrating database: Adding verification_passed column...")
+                cursor.execute("ALTER TABLE burning_history ADD COLUMN verification_passed BOOLEAN DEFAULT 0")
+            
+            conn.commit()
+            print("✅ Database migration complete!")
+    
+    def save_experiment(self, config: Dict, metrics: Dict) -> int:
+        with sqlite3.connect(self.db_path) as conn:
+            cursor = conn.cursor()
+            cursor.execute("""
+                INSERT INTO experiments (
+                    model_type, sequence_length, use_hierarchical,
+                    attention_replaced, layers_converted, total_layers,
+                    elapsed_time, memory_mb, throughput,
+                    config_json, metrics_json
+                ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """, (
+                config.get('model_type'),
+                config.get('sequence_length'),
+                config.get('use_hierarchical'),
+                config.get('attention_replaced'),
+                config.get('layers_converted'),
+                config.get('total_layers'),
+                metrics.get('elapsed_time'),
+                metrics.get('memory_mb'),
+                metrics.get('throughput'),
+                json.dumps(config),
+                json.dumps(metrics)
+            ))
+            conn.commit()
+            return cursor.lastrowid
+    
+    def save_burning(self, burning_info: Dict) -> int:
+        with sqlite3.connect(self.db_path) as conn:
+            cursor = conn.cursor()
+            cursor.execute("""
+                INSERT INTO burning_history (
+                    model_url, output_path, hub_url, use_hierarchical,
+                    dataset_used, conversion_rate, training_steps,
+                    final_loss, evaluation_score, verification_passed
+                ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """, (
+                burning_info.get('model_url'),
+                burning_info.get('output_path'),
+                burning_info.get('hub_url'),
+                burning_info.get('use_hierarchical'),
+                burning_info.get('dataset_used'),
+                burning_info.get('conversion_rate'),
+                burning_info.get('training_steps', 0),
+                burning_info.get('final_loss'),
+                burning_info.get('evaluation_score'),
+                burning_info.get('verification_passed', False),
+            ))
+            conn.commit()
+            return cursor.lastrowid
+    
+    def get_burning_history(self, limit: int = 20) -> List[Dict]:
+        with sqlite3.connect(self.db_path) as conn:
+            conn.row_factory = sqlite3.Row
+            cursor = conn.cursor()
+            cursor.execute("SELECT * FROM burning_history ORDER BY timestamp DESC LIMIT ?", (limit,))
+            return [dict(row) for row in cursor.fetchall()]
+
+
+
+
+
 # =====================================================
 # 모델 버닝 UI 함수 (개선!)
 # =====================================================