Update app.py

app.py

@@ -262,9 +262,9 @@ def infer(genre_txt_content, lyrics_txt_content, num_segments, max_new_tokens):
         logging.info(f"Selected model: {model_path}")
         logging.info(f"Lyrics analysis: {params}")
         
-        # 코러스 섹션 확인
+        # 코러스 섹션 확인 및 로깅
         has_chorus = params['sections']['chorus'] > 0
-        estimated_duration = params.get('estimated_duration', 60)  # 기본값 60초
+        estimated_duration = params.get('estimated_duration', 60)
         
         logging.info(f"Estimated duration: {estimated_duration} seconds")
         logging.info(f"Has chorus sections: {has_chorus}")
@@ -283,7 +283,7 @@ def infer(genre_txt_content, lyrics_txt_content, num_segments, max_new_tokens):
         os.makedirs(output_dir, exist_ok=True)
         empty_output_folder(output_dir)
 
-        # 기본 명령어 구성
+        # 기본 명령어 구성 (지원되는 매개변수만 사용)
         command = [
             "python", "infer.py",
             "--stage1_model", model_path,
@@ -294,16 +294,9 @@ def infer(genre_txt_content, lyrics_txt_content, num_segments, max_new_tokens):
             "--stage2_batch_size", str(config['batch_size']),
             "--output_dir", output_dir,
             "--cuda_idx", "0",
-            "--max_new_tokens", str(actual_max_tokens),
-            "--keep_intermediate"
+            "--max_new_tokens", str(actual_max_tokens)
         ]
 
-        if has_chorus:
-            command.extend([
-                "--segment_duration", str(int(estimated_duration / actual_num_segments)),
-                "--enhance_chorus"
-            ])
-
         # GPU가 있는 경우에만 추가 옵션 적용
         if torch.cuda.is_available():
             command.append("--disable_offload_model")
@@ -348,11 +341,14 @@ def infer(genre_txt_content, lyrics_txt_content, num_segments, max_new_tokens):
         # 결과 처리
         last_mp3 = get_last_mp3_file(output_dir)
         if last_mp3:
-            duration = get_audio_duration(last_mp3)
-            logging.info(f"Generated audio file: {last_mp3}")
-            if duration:
-                logging.info(f"Audio duration: {duration:.2f} seconds")
-                logging.info(f"Expected duration: {estimated_duration} seconds")
+            try:
+                duration = get_audio_duration(last_mp3)
+                logging.info(f"Generated audio file: {last_mp3}")
+                if duration:
+                    logging.info(f"Audio duration: {duration:.2f} seconds")
+                    logging.info(f"Expected duration: {estimated_duration} seconds")
+            except Exception as e:
+                logging.warning(f"Failed to get audio duration: {e}")
             return last_mp3
         else:
             logging.warning("No output audio file generated")