Update app.py

app.py

@@ -6,7 +6,7 @@ import csv
 from datetime import datetime
 from pathlib import Path
 from PIL import Image
-from huggingface_hub import snapshot_download, HfApi # 引入 HfApi 用于直接上传
+from huggingface_hub import CommitScheduler, snapshot_download
 
 # --- 1. 配置区域 ---
 DATASET_REPO_ID = "Emilyxml/moveit"
@@ -15,10 +15,8 @@ LOG_FOLDER = Path("logs")
 LOG_FOLDER.mkdir(parents=True, exist_ok=True)
 TOKEN = os.environ.get("HF_TOKEN")
 
-# 初始化 API 工具
-api = HfApi(token=TOKEN)
-
 # --- 2. 自动下载数据 ---
+# 只有本地为空时才下载，避免每次重启都浪费时间
 if not os.path.exists(DATA_FOLDER) or not os.listdir(DATA_FOLDER):
     try:
         print("🚀 正在从 Dataset 下载数据...")
@@ -33,10 +31,18 @@ if not os.path.exists(DATA_FOLDER) or not os.listdir(DATA_FOLDER):
     except Exception as e:
         print(f"⚠️ 下载失败: {e}")
 
-# --- (注意：这里删除了 CommitScheduler) ---
-# 我们不再需要 Scheduler，改用手动上传
-
-# --- 3. 数据加载 ---
+# --- 3. 恢复后台同步 (解决点击卡顿的关键) ---
+# 使用 Scheduler，提交操作不需要等待网络上传，速度最快
+scheduler = CommitScheduler(
+    repo_id=DATASET_REPO_ID,
+    repo_type="dataset",
+    folder_path=LOG_FOLDER,
+    path_in_repo="logs",
+    every=1, # 每1分钟同步一次，或者是积累了一定数量同步
+    token=TOKEN
+)
+
+# --- 4. 数据加载 ---
 def load_data():
     groups = {}
     if not os.path.exists(DATA_FOLDER):
@@ -75,38 +81,50 @@ def load_data():
 
 ALL_GROUPS, ALL_GROUP_IDS = load_data()
 
-# --- 图片优化函数 ---
-def optimize_image(image_path, max_width=800):
-    if not image_path: return None
+# --- NEW: 更激进的图片优化 (解决加载慢的关键) ---
+def optimize_image(image_path, max_width=500):
+    """
+    调整大小至 500px，对于 User Study 的缩略图查看完全足够。
+    """
+    if not image_path:
+        return None
     try:
         img = Image.open(image_path)
+        # 转换为 RGB 防止 PNG 透明通道在 JPEG 转换时报错
+        if img.mode in ("RGBA", "P"):
+            img = img.convert("RGB")
+            
         if img.width > max_width:
             ratio = max_width / img.width
             new_height = int(img.height * ratio)
+            # 使用 LANCZOS 算法保证缩放质量
             img = img.resize((max_width, new_height), Image.LANCZOS)
         return img
     except Exception as e:
-        print(f"Error: {e}")
+        print(f"Error loading image {image_path}: {e}")
         return None
 
-# --- 4. 核心逻辑 ---
+# --- 5. 核心逻辑 ---
 
 def get_next_question(user_state):
     idx = user_state["index"]
     
+    # 结束逻辑
     if idx >= len(ALL_GROUP_IDS):
         return (
             gr.update(visible=False), gr.update(visible=False), gr.update(visible=False), 
             gr.update(visible=False), gr.update(visible=False), gr.update(visible=False), 
             gr.update(value="## 🎉 测试结束！感谢您的参与。", visible=True),
-            user_state, [] 
+            user_state, []
         )
     
     group_id = ALL_GROUP_IDS[idx]
     group_data = ALL_GROUPS[group_id]
     
-    origin_img = optimize_image(group_data["origin"], max_width=600)
+    # 1. 优化原图
+    origin_img = optimize_image(group_data["origin"], max_width=500)
     
+    # 2. 优化候选图
     candidates = group_data["candidates"].copy()
     random.shuffle(candidates)
     
@@ -116,7 +134,8 @@ def get_next_question(user_state):
     
     for i, path in enumerate(candidates):
         label = f"Option {chr(65+i)}"
-        optimized_img = optimize_image(path, max_width=600)
+        # 优化每张候选图
+        optimized_img = optimize_image(path, max_width=500)
         gallery_items.append((optimized_img, label))
         choices.append(label)
         candidates_info.append({"label": label, "path": path})
@@ -156,43 +175,33 @@ def save_and_next(user_state, candidates_info, selected_options, is_none=False):
                     break
         method_str = "; ".join(selected_methods)
 
-    # --- 关键修改：保存并立即上传 ---
+    # --- 极速保存：只写本地文件，不等待网络上传 ---
     user_filename = f"user_{user_state['user_id']}.csv"
     user_file_path = LOG_FOLDER / user_filename
     
-    # 1. 写入本地 CSV
-    file_exists = user_file_path.exists()
-    with open(user_file_path, "a", newline="", encoding="utf-8") as f:
-        writer = csv.writer(f)
-        if not file_exists:
-            writer.writerow(["user_id", "timestamp", "group_id", "choices", "methods"])
-        writer.writerow([
-            user_state["user_id"], 
-            datetime.now().strftime("%Y-%m-%d %H:%M:%S"), 
-            group_id, 
-            choice_str, 
-            method_str
-        ])
-    
-    # 2. 立即上传到 Dataset (同步操作)
-    try:
-        print(f"Uploading {user_filename} to dataset...")
-        api.upload_file(
-            path_or_fileobj=user_file_path,
-            path_in_repo=f"logs/{user_filename}", # 在 Dataset 中创建 logs 文件夹
-            repo_id=DATASET_REPO_ID,
-            repo_type="dataset"
-        )
-        print("Upload success!")
-    except Exception as e:
-        print(f"⚠️ Upload failed: {e}")
-        # 如果是 Token 权限问题，这里会在 Space Logs 里报错
+    # 使用 Scheduler 提供的锁来保证多线程安全
+    with scheduler.lock:
+        file_exists = user_file_path.exists()
+        with open(user_file_path, "a", newline="", encoding="utf-8") as f:
+            writer = csv.writer(f)
+            if not file_exists:
+                writer.writerow(["user_id", "timestamp", "group_id", "choices", "methods"])
+            writer.writerow([
+                user_state["user_id"], 
+                datetime.now().strftime("%Y-%m-%d %H:%M:%S"), 
+                group_id, 
+                choice_str, 
+                method_str
+            ])
+    
+    print(f"✅ Local Saved: {group_id} (Upload will happen in background)")
             
     user_state["index"] += 1
     return get_next_question(user_state)
 
-# --- 5. 界面构建 ---
+# --- 6. 界面构建 ---
 with gr.Blocks(title="User Study") as demo:
+    
     state_user = gr.State(lambda: {"user_id": str(uuid.uuid4())[:8], "index": 0})
     state_candidates_info = gr.State([]) 
     
@@ -201,6 +210,7 @@ with gr.Blocks(title="User Study") as demo:
     
     with gr.Row():
         with gr.Column(scale=1):
+            # 强制 JPEG 格式，quality 默认 90，显示速度快
             img_origin = gr.Image(label="Reference (参考原图)", interactive=False, height=400, format="jpeg")
         
         with gr.Column(scale=2):
@@ -221,7 +231,4 @@ with gr.Blocks(title="User Study") as demo:
     
     btn_submit.click(fn=lambda s, c, o: save_and_next(s, c, o, is_none=False), inputs=[state_user, state_candidates_info, checkbox_options], outputs=[img_origin, gallery_candidates, checkbox_options, md_instruction, btn_submit, btn_none, md_end, state_user, state_candidates_info])
     
-    btn_none.click(fn=lambda s, c, o: save_and_next(s, c, o, is_none=True), inputs=[state_user, state_candidates_info, checkbox_options], outputs=[img_origin, gallery_candidates, checkbox_options, md_instruction, btn_submit, btn_none, md_end, state_user, state_candidates_info])
-
-if __name__ == "__main__":
-    demo.launch()
+    btn_none.click(fn=lambda s, c, o: save_and_next(s, c, o, is_none=True), inputs=[state_user, state_candidates_info, checkbox