Update app.py

app.py

@@ -5,8 +5,8 @@ import uuid
 import csv
 from datetime import datetime
 from pathlib import Path
-from PIL import Image  # 引入 PIL 用于处理图片
-from huggingface_hub import CommitScheduler, snapshot_download
+from PIL import Image
+from huggingface_hub import snapshot_download, HfApi # 引入 HfApi 用于直接上传
 
 # --- 1. 配置区域 ---
 DATASET_REPO_ID = "Emilyxml/moveit"
@@ -15,6 +15,9 @@ 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 ---
 if not os.path.exists(DATA_FOLDER) or not os.listdir(DATA_FOLDER):
     try:
@@ -30,17 +33,10 @@ if not os.path.exists(DATA_FOLDER) or not os.listdir(DATA_FOLDER):
     except Exception as e:
         print(f"⚠️ 下载失败: {e}")
 
-# --- 3. 启动同步调度器 ---
-scheduler = CommitScheduler(
-    repo_id=DATASET_REPO_ID,
-    repo_type="dataset",
-    folder_path=LOG_FOLDER,
-    path_in_repo="logs",
-    every=1,
-    token=TOKEN
-)
+# --- (注意：这里删除了 CommitScheduler) ---
+# 我们不再需要 Scheduler，改用手动上传
 
-# --- 4. 数据加载 ---
+# --- 3. 数据加载 ---
 def load_data():
     groups = {}
     if not os.path.exists(DATA_FOLDER):
@@ -79,52 +75,38 @@ def load_data():
 
 ALL_GROUPS, ALL_GROUP_IDS = load_data()
 
-# --- NEW: 图片优化函数 (提速关键) ---
+# --- 图片优化函数 ---
 def optimize_image(image_path, max_width=800):
-    """
-    读取图片并调整大小，减少传输时间。
-    max_width: 限制最大宽度为 800px (足够人眼评估)
-    """
-    if not image_path:
-        return None
+    if not image_path: return None
     try:
         img = Image.open(image_path)
-        # 如果图片太大，就缩小
         if img.width > max_width:
             ratio = max_width / img.width
             new_height = int(img.height * ratio)
             img = img.resize((max_width, new_height), Image.LANCZOS)
         return img
     except Exception as e:
-        print(f"Error loading image {image_path}: {e}")
+        print(f"Error: {e}")
         return None
 
-# --- 5. 核心逻辑 ---
+# --- 4. 核心逻辑 ---
 
 def get_next_question(user_state):
-    """准备下一题的数据"""
     idx = user_state["index"]
     
     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(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]
     
-    # 1. 优化原图 (返回 PIL 对象而不是路径)
     origin_img = optimize_image(group_data["origin"], max_width=600)
     
-    # 2. 优化候选图
     candidates = group_data["candidates"].copy()
     random.shuffle(candidates)
     
@@ -134,10 +116,7 @@ def get_next_question(user_state):
     
     for i, path in enumerate(candidates):
         label = f"Option {chr(65+i)}"
-        
-        # 优化每张候选图
         optimized_img = optimize_image(path, max_width=600)
-        
         gallery_items.append((optimized_img, label))
         choices.append(label)
         candidates_info.append({"label": label, "path": path})
@@ -149,11 +128,8 @@ def get_next_question(user_state):
         gr.update(value=gallery_items, visible=True),
         gr.update(choices=choices, value=[], visible=True), 
         gr.update(value=instruction, visible=True),
-        gr.update(visible=True),
-        gr.update(visible=True),
-        gr.update(visible=False),
-        user_state,
-        candidates_info
+        gr.update(visible=True), gr.update(visible=True), gr.update(visible=False),
+        user_state, candidates_info
     )
 
 def save_and_next(user_state, candidates_info, selected_options, is_none=False):
@@ -166,7 +142,6 @@ def save_and_next(user_state, candidates_info, selected_options, is_none=False):
     else:
         if not selected_options:
             raise gr.Error("请至少勾选一个选项，或点击“都不满意”")
-            
         choice_str = "; ".join(selected_options)
         selected_methods = []
         for opt in selected_options: 
@@ -181,27 +156,43 @@ def save_and_next(user_state, candidates_info, selected_options, is_none=False):
                     break
         method_str = "; ".join(selected_methods)
 
-    user_file = LOG_FOLDER / f"user_{user_state['user_id']}.csv"
-    with scheduler.lock:
-        exists = user_file.exists()
-        with open(user_file, "a", newline="", encoding="utf-8") as f:
-            writer = csv.writer(f)
-            if not 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
-            ])
+    # --- 关键修改：保存并立即上传 ---
+    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
             
     user_state["index"] += 1
     return get_next_question(user_state)
 
-# --- 6. 界面构建 ---
+# --- 5. 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([]) 
     
@@ -210,26 +201,15 @@ with gr.Blocks(title="User Study") as demo:
     
     with gr.Row():
         with gr.Column(scale=1):
-            # 将 format 设置为 jpeg 进一步减小体积
             img_origin = gr.Image(label="Reference (参考原图)", interactive=False, height=400, format="jpeg")
         
         with gr.Column(scale=2):
             gallery_candidates = gr.Gallery(
                 label="Candidates (候选结果)", 
-                columns=[2], 
-                height="auto",
-                object_fit="contain",
-                interactive=False,
-                format="jpeg" # 强制输出 JPEG 格式
+                columns=[2], height="auto", object_fit="contain", interactive=False, format="jpeg"
             )
-            
             gr.Markdown("👇 **请在下方勾选您认为最好的结果（可多选）：**")
-            
-            checkbox_options = gr.CheckboxGroup(
-                choices=[], 
-                label="您的选择", 
-                info="对应上方图片的标签 (Option A, B...)"
-            )
+            checkbox_options = gr.CheckboxGroup(choices=[], label="您的选择", info="对应上方图片的标签")
             
             with gr.Row():
                 btn_submit = gr.Button("🚀 提交 (Submit)", variant="primary")
@@ -237,23 +217,11 @@ with gr.Blocks(title="User Study") as demo:
     
     md_end = gr.Markdown(visible=False)
 
-    demo.load(
-        fn=get_next_question,
-        inputs=[state_user],
-        outputs=[img_origin, gallery_candidates, checkbox_options, md_instruction, btn_submit, btn_none, md_end, state_user, state_candidates_info]
-    )
+    demo.load(fn=get_next_question, inputs=[state_user], outputs=[img_origin, gallery_candidates, checkbox_options, md_instruction, btn_submit, btn_none, md_end, state_user, state_candidates_info])
     
-    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_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]
-    )
+    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()