Update app.py

app.py

@@ -3,6 +3,17 @@ from gradio_toggle import Toggle
 import torch
 from huggingface_hub import snapshot_download
 from transformers import pipeline
+import warnings
+warnings.filterwarnings('ignore', category=FutureWarning)
+warnings.filterwarnings('ignore', category=UserWarning)
+
+# sacremoses 설치 확인
+try:
+    import sacremoses
+except ImportError:
+    print("Installing sacremoses...")
+    import subprocess
+    subprocess.check_call(["pip", "install", "sacremoses"])
 
 from xora.models.autoencoders.causal_video_autoencoder import CausalVideoAutoencoder
 from xora.models.transformers.transformer3d import Transformer3DModel
@@ -55,8 +66,14 @@ hf_token = os.getenv("HF_TOKEN")
 openai_api_key = os.getenv("OPENAI_API_KEY")
 client = OpenAI(api_key=openai_api_key)
 
-# Initialize translation pipeline
-translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en")
+# Initialize translation pipeline with device and clean_up settings
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+translator = pipeline(
+    "translation", 
+    model="Helsinki-NLP/opus-mt-ko-en",
+    device=device,
+    clean_up_tokenization_spaces=True
+)
 
 # Korean text detection function
 def contains_korean(text):
@@ -196,14 +213,14 @@ pipeline = XoraVideoPipeline(
 ).to(device)
 
 
-# txt2vid와 img2vid의 현재 상태 값들을 수정
-txt2vid_current_height = gr.State(value=320)  # 수정됨
-txt2vid_current_width = gr.State(value=512)   # 수정됨
-txt2vid_current_num_frames = gr.State(value=257)  # 10.3초
+# 현재 선택된 값들을 저장할 상태 변수들
+txt2vid_current_height = gr.State(value=320)     # 수정됨
+txt2vid_current_width = gr.State(value=512)      # 수정됨
+txt2vid_current_num_frames = gr.State(value=257) # 10.3초
 
-img2vid_current_height = gr.State(value=320)  # 수정됨
-img2vid_current_width = gr.State(value=512)   # 수정됨
-img2vid_current_num_frames = gr.State(value=257)  # 10.3초
+img2vid_current_height = gr.State(value=320)     # 수정됨
+img2vid_current_width = gr.State(value=512)      # 수정됨
+img2vid_current_num_frames = gr.State(value=257) # 10.3초
 
 # Preset options for resolution and frame configuration
 # Convert frames to seconds assuming 25 FPS
@@ -521,10 +538,11 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
 
                     txt2vid_preset = gr.Dropdown(
                         choices=[p["label"] for p in preset_options],
-                        value="512x320, 10.3초",
+                        value="[16:9] 512x320, 10.3초",  # 정확한 라벨로 수정
                         label="Step 2: 해상도 프리셋 선택",
                     )
 
+
                     txt2vid_frame_rate = gr.Slider(
                         label="Step 3: 프레임 레이트",
                         minimum=21,
@@ -579,9 +597,11 @@ with gr.Blocks(theme=gr.themes.Soft()) as iface:
 
                     img2vid_preset = gr.Dropdown(
                         choices=[p["label"] for p in preset_options],
-                        value="512x320, 10.3초",
+                        value="[16:9] 512x320, 10.3초",  # 정확한 라벨로 수정
                         label="Step 3: 해상도 프리셋 선택",
-                    )
+                    )                    
+                    
+
 
                     img2vid_frame_rate = gr.Slider(
                         label="Step 4: 프레임 레이트",