Update app.py

app.py

@@ -7,7 +7,7 @@ from datetime import datetime
 from functools import partial
 import json
 import io
-from huggingface_hub import HfApi # HfFolder 未使用
+from huggingface_hub import HfApi
 from huggingface_hub.hf_api import HfHubHTTPError
 import traceback
 
@@ -31,13 +31,13 @@ if os.path.exists(BASE_IMAGE_DIR):
     except Exception as e: print(f"错误：在扫描 '{BASE_IMAGE_DIR}' 时发生错误: {e}"); METHOD_ROOTS = []
 else: print(f"警告：基础目录 '{BASE_IMAGE_DIR}' 不存在。将无法加载候选图片。")
 
-SUBJECTS = ["subj01", "subj02", "subj05", "subj07"]
+SUBJECTS = ["subj01", "subj02", "subj05", "subj07"] 
 SENTINEL_TRIAL_INTERVAL = 20
 NUM_TRIALS_PER_RUN = 100 
 
 DATASET_REPO_ID = "YanmHa/image-aligned-experiment-data"
-INDIVIDUAL_LOGS_FOLDER = "individual_choice_logs"
-BATCH_LOG_FOLDER = "run_logs_batch" 
+INDIVIDUAL_LOGS_FOLDER = "individual_choice_logs" # 单个日志文件夹（如果将来恢复单条保存，此名称仍可用）
+BATCH_LOG_FOLDER = "run_logs_batch" # 用于批量保存的文件夹
 CSS = ".gr-block {margin-top: 4px !important; margin-bottom: 4px !important;} .compact_button { padding: 4px 8px; min-width: auto; }"
 
 # ==== 加载所有可用的目标图片 ====
@@ -57,7 +57,7 @@ if not master_image_list: print(f"关键错误：由于 '{TARGET_DIR}' 问题，
 
 # ==== 辅助函数 ====
 def get_next_trial_info(current_trial_idx_in_run, current_run_image_list_for_trial, num_trials_in_this_run_for_trial):
-    # ... (此函数与上一版几乎一致，确保 left_display_label 和 right_display_label 是匿名的) ...
+    # ... (此函数与您上一版代码完全一致, 返回匿名显示标签和内部详细标签) ...
     global TARGET_DIR, METHOD_ROOTS, SUBJECTS, SENTINEL_TRIAL_INTERVAL
     if not current_run_image_list_for_trial or current_trial_idx_in_run >= num_trials_in_this_run_for_trial: return None, current_trial_idx_in_run
     img_filename_original = current_run_image_list_for_trial[current_trial_idx_in_run]
@@ -81,8 +81,6 @@ def get_next_trial_info(current_trial_idx_in_run, current_run_image_list_for_tri
             trial_info["is_sentinel"] = True
             sentinel_candidate_target_tuple = ("目标图像", target_full_path)
             random_reconstruction_candidate_tuple = random.choice(pool)
-            # candidates_for_sentinel 的每个元素是 ((display_label, path_for_display), (internal_label, path_for_internal))
-            # 但为了简化，display_label 可以简单，internal_label 存详细信息
             candidates_for_sentinel = [
                 (("目标图像", target_full_path), sentinel_candidate_target_tuple[0]), 
                 (("重建图", random_reconstruction_candidate_tuple[1]), random_reconstruction_candidate_tuple[0])
@@ -102,38 +100,79 @@ def get_next_trial_info(current_trial_idx_in_run, current_run_image_list_for_tri
         })
     return trial_info, current_trial_idx_in_run + 1
 
-def save_single_log_to_hf_dataset(log_entry, user_identifier_str):
-    # ... (此函数与您上一版代码完全一致) ...
-    global DATASET_REPO_ID, INDIVIDUAL_LOGS_FOLDER
-    if not isinstance(log_entry, dict): print(f"错误：单个日志条目不是字典格式，无法保存：{log_entry}"); return
-    current_user_id = user_identifier_str if user_identifier_str else "unknown_user_session"
-    identifier_safe = str(current_user_id).replace('.', '_').replace(':', '_').replace('/', '_').replace(' ', '_') # 添加替换空格
-    print(f"用户 {identifier_safe} - 保存日志 for image {log_entry.get('image_id', 'Unknown')}...")
+# --- (原 save_single_log_to_hf_dataset 函数可以删除或注释掉，因为不再每次选择都调用) ---
+# def save_single_log_to_hf_dataset(log_entry, user_identifier_str):
+#     ...
+
+# --- 新增/修改：批量保存累积日志的函数 ---
+def save_collected_logs_batch(list_of_log_entries, user_identifier_str, current_run_no_for_filename):
+    global DATASET_REPO_ID, BATCH_LOG_FOLDER # 使用 BATCH_LOG_FOLDER
+    if not list_of_log_entries:
+        print("没有累积的日志可供批量保存。")
+        return True # 没有东西要保存，也算“成功”
+
+    identifier_safe = str(user_identifier_str if user_identifier_str else "unknown_user_session").replace('.', '_').replace(':', '_').replace('/', '_').replace(' ', '_')
+    print(f"用户 {identifier_safe} - 准备批量保存截至轮次 {current_run_no_for_filename} 的 {len(list_of_log_entries)} 条日志...")
+    
     try:
         token = os.getenv("HF_TOKEN")
-        if not token: print("错误：HF_TOKEN 未设置。无法保存日志。"); return
-        if not DATASET_REPO_ID: print("错误：DATASET_REPO_ID 未配置。"); return
+        if not token:
+            print("错误：HF_TOKEN 未设置。无法批量保存日志。")
+            return False # 指示保存失败
+        if not DATASET_REPO_ID:
+            print("错误：DATASET_REPO_ID 未配置。无法批量保存日志。")
+            return False # 指示保存失败
+
         api = HfApi(token=token)
-        image_id_safe = os.path.splitext(log_entry.get("image_id", "unk_img"))[0].replace('.', '_').replace(':', '_').replace('/', '_')
-        ts = datetime.now().strftime('%Y%m%d_%H%M%S_%f')
-        fname = f"run{log_entry.get('run_no','X')}_trial{log_entry.get('trial_sequence_in_run','Y')}_img{image_id_safe}_{ts}.json"
-        path_in_repo = f"{INDIVIDUAL_LOGS_FOLDER}/{identifier_safe}/{fname}"
-        json_content = json.dumps(log_entry, ensure_ascii=False, indent=2)
-        log_bytes = json_content.encode('utf-8')
-        f_obj = io.BytesIO(log_bytes)
-        print(f"准备上传: {path_in_repo} ({len(log_bytes)} bytes)")
-        api.upload_file(path_or_fileobj=f_obj,path_in_repo=path_in_repo,repo_id=DATASET_REPO_ID,repo_type="dataset",
-                        commit_message=f"Log: img {log_entry.get('image_id','N/A')}, run {log_entry.get('run_no','N/A')} by {identifier_safe}")
-        print(f"日志成功保存到: {DATASET_REPO_ID}/{path_in_repo}")
-    except Exception as e: print(f"保存日志失败 (img {log_entry.get('image_id','Unk')}, user {identifier_safe}): {e}"); traceback.print_exc()
+        
+        timestamp_str = datetime.now().strftime('%Y%m%d_%H%M%S_%f')
+        # 文件名可以指明这是哪一“批”的日志，例如基于当前的轮次号
+        batch_filename = f"batch_user-{identifier_safe}_upto-run{current_run_no_for_filename}_{timestamp_str}.jsonl"
+        # 路径结构: BATCH_LOG_FOLDER / USER_SESSION_ID_SAFE / FILENAME.jsonl
+        path_in_repo = f"{BATCH_LOG_FOLDER}/{identifier_safe}/{batch_filename}" 
+
+        jsonl_content = ""
+        for log_entry in list_of_log_entries:
+            try:
+                if isinstance(log_entry, dict):
+                    jsonl_content += json.dumps(log_entry, ensure_ascii=False) + "\n"
+                else:
+                    print(f"警告：批量保存时，日志条目不是字典格式，跳过：{log_entry}")
+            except Exception as json_err:
+                print(f"错误：批量保存时，序列化单条日志时出错: {log_entry}. 错误: {json_err}")
+                jsonl_content += json.dumps({"error": "serialization_failed_in_batch", 
+                                             "original_data_preview": str(log_entry)[:100],
+                                             "timestamp": datetime.now().isoformat()}, ensure_ascii=False) + "\n"
+        
+        if not jsonl_content.strip():
+            print(f"用户 {identifier_safe} 截至轮次 {current_run_no_for_filename} 没有可序列化的日志内容。")
+            return True # 没有东西要上传
 
+        log_bytes = jsonl_content.encode('utf-8')
+        file_like_object = io.BytesIO(log_bytes)
+
+        print(f"准备批量上传日志文件: {path_in_repo} ({len(log_bytes)} bytes)")
+        api.upload_file(
+            path_or_fileobj=file_like_object,
+            path_in_repo=path_in_repo,
+            repo_id=DATASET_REPO_ID,
+            repo_type="dataset", # 确保与您的DATASET_REPO_ID类型匹配
+            commit_message=f"Batch logs for user {identifier_safe} up to run {current_run_no_for_filename}"
+        )
+        print(f"批量日志已成功保存到 HF Dataset: {DATASET_REPO_ID}/{path_in_repo}")
+        return True # 指示保存成功
+    except Exception as e:
+        print(f"批量保存日志 (user {identifier_safe}, up to run {current_run_no_for_filename}) 到 Hugging Face Dataset 时发生严重错误: {e}")
+        traceback.print_exc()
+        return False # 指示保存失败
+
+# ==== 主要的 Gradio 事件处理函数 ====
 def process_experiment_step(
     s_trial_idx_val, s_run_no_val, s_user_logs_val, s_current_trial_data_val, s_user_session_id_val,
     s_current_run_image_list_val, s_num_trials_this_run_val,
-    action_type=None, choice_value=None, request: gr.Request = None):
-    # ... (此函数与上一版几乎一致，确保所有yield/return都返回正确数量的值) ...
-    # ... (并且 chosen_raw_label 现在使用 chosen_internal_label, left/right_method_subject_filename 使用 internal_label) ...
-    global master_image_list, NUM_TRIALS_PER_RUN, outputs_ui_components_definition 
+    action_type=None, choice_value=None, request: gr.Request = None
+):
+    global master_image_list, NUM_TRIALS_PER_RUN, outputs_ui_components_definition
     output_s_trial_idx = s_trial_idx_val; output_s_run_no = s_run_no_val
     output_s_user_logs = list(s_user_logs_val); output_s_current_trial_data = dict(s_current_trial_data_val) if s_current_trial_data_val else {}
     output_s_user_session_id = s_user_session_id_val; output_s_current_run_image_list = list(s_current_run_image_list_val)
@@ -155,42 +194,91 @@ def process_experiment_step(
                 if len(parts) == 3: parsed_chosen_method, parsed_chosen_subject, parsed_chosen_filename = parts[0].strip(), parts[1].strip(), parts[2].strip()
                 elif len(parts) == 2: parsed_chosen_method, parsed_chosen_subject = parts[0].strip(), parts[1].strip()
                 elif len(parts) == 1: parsed_chosen_method = parts[0].strip()
-            log_entry = {"timestamp": datetime.now().isoformat(), "user_identifier": user_identifier_for_logging, "run_no": output_s_run_no, "image_id": output_s_current_trial_data["data"]["image_id"],
-                         "left_internal_label": output_s_current_trial_data["data"]["left_internal_label"], "right_internal_label": output_s_current_trial_data["data"]["right_internal_label"],
-                         "chosen_side": choice_value, "chosen_internal_label": chosen_internal_label, "chosen_method": parsed_chosen_method, "chosen_subject": parsed_chosen_subject, "chosen_filename": parsed_chosen_filename,
-                         "trial_sequence_in_run": output_s_current_trial_data["data"]["cur_no"], "is_sentinel": output_s_current_trial_data["data"]["is_sentinel"]}
-            output_s_user_logs.append(log_entry)
-            print(f"用户 {user_identifier_for_logging} 记录选择...准备立即保存...")
-            save_single_log_to_hf_dataset(log_entry, user_identifier_for_logging)
+            log_entry = { # ... (log_entry 创建与之前一致) ...
+                "timestamp": datetime.now().isoformat(), "user_identifier": user_identifier_for_logging, "run_no": output_s_run_no, 
+                "image_id": output_s_current_trial_data["data"]["image_id"],
+                "left_internal_label": output_s_current_trial_data["data"]["left_internal_label"], 
+                "right_internal_label": output_s_current_trial_data["data"]["right_internal_label"],
+                "chosen_side": choice_value, "chosen_internal_label": chosen_internal_label, 
+                "chosen_method": parsed_chosen_method, "chosen_subject": parsed_chosen_subject, "chosen_filename": parsed_chosen_filename,
+                "trial_sequence_in_run": output_s_current_trial_data["data"]["cur_no"], 
+                "is_sentinel": output_s_current_trial_data["data"]["is_sentinel"]
+            }
+            output_s_user_logs.append(log_entry) # <--- 累积日志到列表
+            print(f"用户 {user_identifier_for_logging} 记录选择 (img: {log_entry['image_id']})。当前批次日志数: {len(output_s_user_logs)}")
+            # !!! 移除 save_single_log_to_hf_dataset 的调用 !!!
         else:
+            # ... (错误处理) ...
             print(f"用户 {user_identifier_for_logging} 错误：记录选择时数据为空！")
             error_ui_updates = create_ui_error_tuple("记录选择时内部错误。", f"用户ID: {user_id_display_text} | 进度：{output_s_trial_idx}/{output_s_num_trials_this_run}")
             return output_s_trial_idx, output_s_run_no, output_s_user_logs, output_s_current_trial_data, output_s_user_session_id, output_s_current_run_image_list, output_s_num_trials_this_run, *error_ui_updates
+
     if action_type == "start_experiment":
         is_first = (output_s_num_trials_this_run == 0 and output_s_trial_idx == 0 and output_s_run_no == 1)
-        is_completed = (output_s_num_trials_this_run > 0 and output_s_trial_idx >= output_s_num_trials_this_run)
-        if is_first or is_completed:
-            if not master_image_list: error_ui = create_ui_error_tuple("错误: 无可用目标图片!", f"用户ID: {user_id_display_text} | 进度: 0/0"); return 0, output_s_run_no, [], {}, output_s_user_session_id, [], 0, *error_ui
-            if is_completed: output_s_run_no += 1
+        is_completed_for_restart = (output_s_num_trials_this_run > 0 and output_s_trial_idx >= output_s_num_trials_this_run)
+        
+        if is_first or is_completed_for_restart:
+            # ... (开始新轮次的逻辑，与之前代码一致，例如：随机选择图片，重置trial_idx等) ...
+            # ... (注意：当 is_completed_for_restart 时，output_s_run_no 应该已经增加了) ...
+            # output_s_user_logs 不在这里重置，它会累积
+            if not master_image_list: error_ui = create_ui_error_tuple("错误: 无可用目标图片!", f"用户ID: {user_id_display_text} | 进度: 0/0"); return 0, output_s_run_no, output_s_user_logs, {}, output_s_user_session_id, [], 0, *error_ui # output_s_user_logs is passed as is
+            
+            # 这段逻辑用于确定是否是真正的“下一轮”并增加轮次号
+            # 如果是点击“开始实验”按钮且上一轮刚结束，则轮次号增加
+            if is_completed_for_restart: 
+                # 此时 output_s_trial_idx 指向的是刚完成轮次的总数，output_s_run_no 是刚完成的轮次号
+                # 所以，下一轮的轮次号应该是 output_s_run_no + 1
+                # 但 process_experiment_step 的状态管理是，它返回的值会成为下一次调用的输入值
+                # 所以，如果这里要开始新轮次，应该在这里增加 output_s_run_no
+                 output_s_run_no += 1 # 增加轮次号为新的一轮
+
             num_avail = len(master_image_list); run_size = min(num_avail, NUM_TRIALS_PER_RUN)
-            if run_size == 0: error_ui = create_ui_error_tuple("错误: 采样图片数为0!", f"用户ID: {user_id_display_text} | 进度: 0/0"); return 0, output_s_run_no, [], {}, output_s_user_session_id, [], 0, *error_ui
+            if run_size == 0: error_ui = create_ui_error_tuple("错误: 采样图片数为0!", f"用户ID: {user_id_display_text} | 进度: 0/0"); return 0, output_s_run_no, output_s_user_logs, {}, output_s_user_session_id, [], 0, *error_ui
+            
             output_s_current_run_image_list = random.sample(master_image_list, run_size)
-            output_s_num_trials_this_run = run_size; output_s_trial_idx = 0; output_s_user_logs = []; output_s_current_trial_data = {}
-            # 新的 user_session_id (即 participant_id) 在欢迎页已收集并传入 s_user_session_id_val
-            # 这里不再重新生成基于时间戳的 output_s_user_session_id，而是使用已有的。
-            # user_identifier_for_logging 已经使用了 output_s_user_session_id
+            output_s_num_trials_this_run = run_size
+            output_s_trial_idx = 0 # 新轮次从0开始
+            # output_s_user_logs 不重置
+            output_s_current_trial_data = {}
+            
+            # 只有在会话ID为空（即应用刚启动，用户第一次点击“同意”后，再点击“开始实验”）时才生成新的基于时间戳的ID
+            # 或者，如果策略是每轮实验都用新的用户ID，则可以在这里（is_first or is_completed_for_restart）重新生成
+            # 当前代码中，user_session_id 是在欢迎页的 handle_agree_and_start 中设置的，这里不再修改它
+            # user_identifier_for_logging 会使用这个ID
             print(f"开始/继续轮次 {output_s_run_no} (用户ID: {output_s_user_session_id}). 随机选择 {output_s_num_trials_this_run} 张图片.")
-        else:
+
+        else: # 在轮次中途点击了开始按钮
             print(f"用户 {user_identifier_for_logging} 在第 {output_s_run_no} 轮，试验 {output_s_trial_idx} 点击开始，但轮次未完成。忽略。")
             no_change_ui = create_no_change_tuple()
             return output_s_trial_idx, output_s_run_no, output_s_user_logs, output_s_current_trial_data, output_s_user_session_id, output_s_current_run_image_list, output_s_num_trials_this_run, *no_change_ui
+
+    # --- 轮次结束处理 ---
     if output_s_trial_idx >= output_s_num_trials_this_run and output_s_num_trials_this_run > 0:
-        if output_s_user_logs: print(f"用户 {output_s_user_session_id} 完成第 {output_s_run_no} 轮。")
+        logs_for_this_run_count = sum(1 for log in output_s_user_logs if log.get("run_no") == output_s_run_no) # 粗略计算本轮日志数
+        if logs_for_this_run_count > 0 or not output_s_user_logs: # 如果本轮有日志，或总日志为空（可能刚开始）
+             print(f"用户 {output_s_user_session_id} 完成第 {output_s_run_no} 轮。累积日志数: {len(output_s_user_logs)}")
+
+        # 检查是否达到批量保存的条件
+        if output_s_run_no > 0 and output_s_run_no % 10 == 0:
+            if output_s_user_logs: # 只有当有日志时才尝试保存
+                print(f"达到第 {output_s_run_no} 轮，准备批量保存累积的日志...")
+                save_success = save_collected_logs_batch(list(output_s_user_logs), user_identifier_for_logging, output_s_run_no)
+                if save_success:
+                    print("批量日志已成功（或尝试）保存，将清空累积日志列表。")
+                    output_s_user_logs = [] # 清空已保存的日志
+                else:
+                    print("警告：批量日志保存失败。日志将继续累积。")
+            else:
+                print(f"第 {output_s_run_no} 轮结束，但累积日志为空，无需批量保存。")
+        
         prog_text = f"用户ID: {output_s_user_session_id} | 进度：{output_s_num_trials_this_run}/{output_s_num_trials_this_run} | 第 {output_s_run_no} 轮 🎉"
-        ui_updates = list(create_ui_error_tuple(f"🎉 第 {output_s_run_no} 轮完成！请点击“开始试验 / 下一轮”。", prog_text))
+        ui_updates = list(create_ui_error_tuple(f"🎉 第 {output_s_run_no} 轮完成！请点击“开始试验 / 下一轮”开始新的轮次。", prog_text))
         ui_updates[7]=gr.update(interactive=True); ui_updates[8]=gr.update(interactive=False); ui_updates[9]=gr.update(interactive=False)
         ui_updates[0]=gr.update(value=None,visible=False); ui_updates[1]=gr.update(value=None,visible=False); ui_updates[2]=gr.update(value=None,visible=False)
         yield output_s_trial_idx, output_s_run_no, output_s_user_logs, output_s_current_trial_data, output_s_user_session_id, output_s_current_run_image_list, output_s_num_trials_this_run, *ui_updates; return
+    
+    # --- 获取并显示下一个试验 ---
+    # ... (与上一版代码相同，确保 prog_text 使用了新的 user_id_display_text 或 output_s_user_session_id) ...
     if not output_s_current_run_image_list or output_s_num_trials_this_run == 0:
         error_ui = create_ui_error_tuple("错误: 无法加载试验图片 (列表为空)", f"用户ID: {user_id_display_text} | 进度: N/A")
         return output_s_trial_idx, output_s_run_no, output_s_user_logs, {"data": None}, output_s_user_session_id, [], 0, *error_ui
@@ -209,20 +297,17 @@ def process_experiment_step(
     time.sleep(3)
     ui_show_candidates_updates = list(create_no_change_tuple())
     ui_show_candidates_updates[0]=gr.update(value=None,visible=False); ui_show_candidates_updates[1]=gr.update(value=trial_info["left_path"],visible=True); ui_show_candidates_updates[2]=gr.update(value=trial_info["right_path"],visible=True)
-    ui_show_candidates_updates[3]=gr.update(value=trial_info["left_display_label"], visible=False) # 隐藏标签
-    ui_show_candidates_updates[4]=gr.update(value=trial_info["right_display_label"], visible=False) # 隐藏标签
+    ui_show_candidates_updates[3]=gr.update(value=trial_info["left_display_label"], visible=False); ui_show_candidates_updates[4]=gr.update(value=trial_info["right_display_label"], visible=False)
     ui_show_candidates_updates[5]="请选择更像原图的一张"; ui_show_candidates_updates[6]=prog_text
     ui_show_candidates_updates[7]=gr.update(interactive=False); ui_show_candidates_updates[8]=gr.update(interactive=True); ui_show_candidates_updates[9]=gr.update(interactive=True)
     yield next_s_trial_idx_for_state, output_s_run_no, output_s_user_logs, output_s_current_trial_data, output_s_user_session_id, output_s_current_run_image_list, output_s_num_trials_this_run, *ui_show_candidates_updates
 
-# ==== Gradio UI 定义 ====
+
 welcome_page_markdown = """
 ## 欢迎加入实验！
 您好！非常感谢您抽出宝贵时间参与我们的视觉偏好评估实验。您的选择将帮助我们改进重建算法，让机器生成的图像更贴近人类视觉体验!
-
 1.  **实验目的**
     通过比较两幅 重建图像 与原始 目标图像 的相似度。
-
 2.  **操作流程**
     * 点击下方的「我已阅读并同意开始实验」按钮。
     * 然后点击主实验界面的「开始试验 / 下一轮」按钮。
@@ -230,212 +315,123 @@ welcome_page_markdown = """
     * 随后自动切换到 **两张重建图像**。
     * 根据刚才的观察记忆，选出您认为与目标图像最相似的一张。
     * 选择后系统会自动进入下一轮比较。
-
 3.  **温馨提示**
     * 请勿刷新或关闭页面，以免中断实验。
     * 若图片加载稍有延迟，请耐心等待；持续异常可联系邮箱 yangminghan@bupt.edu.cn。
     * 本实验将保护您的任何个人隐私信息，所有数据仅用于学术研究，请您认真选择和填写。
-
 4.  **奖励说明**
     * 完成全部轮次后，请截图记录您所完成的实验总数（可累积,页面左下角将显示进度,请保证截取到为您分配的ID，轮次）。
     * 将截图发送至邮箱 yangminghan@bupt.edu.cn，我们将在核验后发放奖励。
-
 再次感谢您的参与与支持！您每一次认真选择都对我们的研究意义重大。祝您一切顺利，实验愉快！
 """
 
-# 新增：处理同意并开始实验的函数
 def handle_agree_and_start(name, gender, age, education, request: gr.Request):
-    # 简单验证
     error_messages_list = []
-    if not name or str(name).strip() == "":
-        error_messages_list.append("姓名 不能为空。")
-    
-    # 对于 Radio 和 Dropdown，如果初始没有值且用户未选择，它们可能返回 None
-    if gender is None or str(gender).strip() == "": # Radio 可能返回None或空字符串
-        error_messages_list.append("性别 必须选择。")
-    
-    # 对于 gr.Number，如果未填写，它可能是 None。age=0 可能也是无效的，取决于您的逻辑
-    if age is None: # 或者 age == 0 如果0是无效年龄
-        error_messages_list.append("年龄 不能为空或为0。")
-    elif not (isinstance(age, (int, float)) and 1 <= age <= 120): # 假设年龄范围1-120
-        try: # 尝试转换以处理从UI传来的可能是字符串的数字
-            num_age = float(age)
-            if not (1 <= num_age <= 120):
-                 error_messages_list.append("年龄必须在 1 到 120 之间。")
-        except (ValueError, TypeError):
-            error_messages_list.append("年龄必须是一个有效的数字。")
-
-    if education is None or str(education).strip() == "其他": # Dropdown 可能返回None或空字符串
-        error_messages_list.append("学历 必须选择。")
+    if not name or str(name).strip() == "": error_messages_list.append("姓名 不能为空。")
+    if gender is None or str(gender).strip() == "": error_messages_list.append("性别 必须选择。")
+    if age is None: error_messages_list.append("年龄 不能为空。")
+    elif not (isinstance(age, (int, float)) and 1 <= age <= 120):
+        try: num_age = float(age);
+        except (ValueError, TypeError): error_messages_list.append("年龄必须是一个有效的数字。")
+        else: 
+            if not (1 <= num_age <= 120): error_messages_list.append("年龄必须在 1 到 120 之间。")
+    if education is None or str(education).strip() == "" or str(education).strip() == "其他" and not name: # 稍微修改逻辑，如果选了“其他”但没填姓名，也可能需要提示
+        # 更简单的逻辑：如果education是None或空字符串
+        if education is None or str(education).strip() == "":
+             error_messages_list.append("学历 必须选择。")
 
     if error_messages_list:
         full_error_message = "请修正以下错误：\n" + "\n".join([f"- {msg}" for msg in error_messages_list])
         print(f"用户输入验证失败: {full_error_message}")
-        # 返回更新，以显示错误并保持在欢迎页面
-        # 对应 outputs: [s_user_session_id, s_show_experiment_ui, welcome_container, experiment_container, welcome_error_msg]
-        return (
-            gr.update(),                 # s_user_session_id (不改变)
-            False,                       # s_show_experiment_ui (保持为False，显示欢迎页)
-            gr.update(visible=True),     # welcome_container (保持可见)
-            gr.update(visible=False),    # experiment_container (保持隐藏)
-            full_error_message           # welcome_error_msg (显示错误信息)
-        )
+        return gr.update(), False, gr.update(visible=True), gr.update(visible=False), full_error_message
             
-    # 清理和连接信息作为用户ID
     s_name = str(name).strip().replace(" ","_").replace("/","_").replace("\\","_")
     s_gender = str(gender).strip().replace(" ","_").replace("/","_").replace("\\","_")
-    s_age = str(int(float(age))) # 确保是整数再转字符串
+    s_age = str(int(float(age))) 
     s_education = str(education).strip().replace(" ","_").replace("/","_").replace("\\","_")
-
     user_id_str = f"N-{s_name}_G-{s_gender}_A-{s_age}_E-{s_education}"
-        
     print(f"用户信息收集完毕，生成用户ID: {user_id_str}")
-        
-    # 成功，切换界面并清空错误信息
     return user_id_str, True, gr.update(visible=False), gr.update(visible=True), ""
 
-
 with gr.Blocks(css=CSS, title="图像重建主观评估") as demo:
-    # --- 所有 State 变量首先定义 ---
     s_show_experiment_ui = gr.State(False)
-    s_trial_index = gr.State(0)
-    s_run_no = gr.State(1)
-    s_user_logs = gr.State([])
-    s_current_trial_data = gr.State({})
-    s_user_session_id = gr.State(None) # 将存储由用户信息构成的ID
-    s_current_run_image_list = gr.State([])
-    s_num_trials_this_run = gr.State(0)
+    s_trial_index = gr.State(0); s_run_no = gr.State(1); s_user_logs = gr.State([])
+    s_current_trial_data = gr.State({}); s_user_session_id = gr.State(None)
+    s_current_run_image_list = gr.State([]); s_num_trials_this_run = gr.State(0)
 
-    # --- UI 容器 ---
     welcome_container = gr.Column(visible=True) 
     experiment_container = gr.Column(visible=False)
 
-    # --- 欢迎页面 UI ---
     with welcome_container:
-        gr.Markdown(welcome_page_markdown) # welcome_page_markdown 在全局定义
+        gr.Markdown(welcome_page_markdown)
         with gr.Row():
-            user_name_input = gr.Textbox(label="请输入您的姓名", placeholder="例如：张三 ")
-            user_gender_input = gr.Radio(label="性别", choices=["男", "女"])
+            user_name_input = gr.Textbox(label="请输入您的姓名或代号 (例如 ZS 或 User001)", placeholder="例如：张三 -> ZS") # 修改了label
+            user_gender_input = gr.Radio(label="性别", choices=["男", "女", "其他", "不愿透露"], value="不愿透露")
         with gr.Row():
-            user_age_input = gr.Number(label="年龄 (请输入整数)", minimum=1, maximum=120, step=1)
-            user_education_input = gr.Dropdown(
-                label="学历", 
-                choices=["其他","初中及以下","高中（含在读）", "大专（含在读）", "本科（含在读）", "硕士（含在读）", "博士（含在读）"]
-            )
-        welcome_error_msg = gr.Markdown(value="") # 用于显示欢迎页的错误信息
+            user_age_input = gr.Number(label="年龄 (请输入1-120的整数)", minimum=1, maximum=120, step=1, value=25) # 修改了label和value
+            user_education_input = gr.Dropdown(label="学历", choices=["其他","初中及以下","高中（含中专）", "大专", "本科", "硕士", "博士（含在读）"], value="本科") # 修改了选项和value
+        welcome_error_msg = gr.Markdown(value="")
         btn_agree_and_start = gr.Button("我已阅读上述说明并同意参与实验")
 
-    # --- 实验主界面 UI ---
     with experiment_container:
         gr.Markdown("## 🧠 图像重建主观评估实验")
-        gr.Markdown(
-            f"每轮实验大约有 {NUM_TRIALS_PER_RUN} 次比较。" # 移除了原来对流程的详细描述，因为已在欢迎页说明
-        )
-        # 候选图片区域的新布局
+        gr.Markdown(f"每轮实验大约有 {NUM_TRIALS_PER_RUN} 次比较。")
         with gr.Row():
-            with gr.Column(scale=1, min_width=300): # 左边列
+            with gr.Column(scale=1, min_width=300):
                 left_img = gr.Image(label="左候选图", visible=False, height=400, interactive=False)
-                # left_lbl Textbox 现在将被永久隐藏，但保留以匹配输出数量
                 left_lbl = gr.Textbox(label="左图信息", visible=False, interactive=False, max_lines=1) 
                 btn_left = gr.Button("选择左图 (更相似)", interactive=False, elem_classes="compact_button")
-            
-            with gr.Column(scale=1, min_width=300): # 右边列
+            with gr.Column(scale=1, min_width=300):
                 right_img = gr.Image(label="右候选图", visible=False, height=400, interactive=False)
-                # right_lbl Textbox 现在将被永久隐藏
                 right_lbl = gr.Textbox(label="右图信息", visible=False, interactive=False, max_lines=1)
                 btn_right = gr.Button("选择右图 (更相似)", interactive=False, elem_classes="compact_button")
-
-        # 目标图像单独一行，跨列显示或居中
-        with gr.Row():
-            target_img = gr.Image(label="目标图像 (观察3秒后消失)", visible=False, height=400, interactive=False)
-        
-        # 状态和进度文本
-        with gr.Row():
-            status_text = gr.Markdown(value="请点击“开始试验 / 下一轮”按钮。")
-        with gr.Row():
-            progress_text = gr.Markdown() # 用于显示 用户ID | 进度 | 轮次
-        
-        # 开始按钮单独一行
-        with gr.Row():
-            btn_start = gr.Button("开始试验 / 下一轮")
-
-        # 隐藏的 File 组件，用于匹配输出数量 (如果仍需要)
+        with gr.Row(): target_img = gr.Image(label="目标图像 (观察3秒后消失)", visible=False, height=400, interactive=False)
+        with gr.Row(): status_text = gr.Markdown(value="请点击“开始试验 / 下一轮”按钮。")
+        with gr.Row(): progress_text = gr.Markdown()
+        with gr.Row(): btn_start = gr.Button("开始试验 / 下一轮")
         file_out_placeholder = gr.File(label=" ", visible=False, interactive=False)
 
-
-    # --- 定义UI组件列表，用于简化事件处理函数的输出 ---
     outputs_ui_components_definition = [ 
-        target_img, left_img, right_img,
-        left_lbl, right_lbl, # 即使隐藏，也包含在列表以维持原函数的输出结构
-        status_text, progress_text,
-        btn_start, btn_left, btn_right,
-        file_out_placeholder 
+        target_img, left_img, right_img, left_lbl, right_lbl, status_text, progress_text,
+        btn_start, btn_left, btn_right, file_out_placeholder 
     ]
-    
     click_inputs_base = [ 
         s_trial_index, s_run_no, s_user_logs, s_current_trial_data, s_user_session_id,
         s_current_run_image_list, s_num_trials_this_run
     ]
     event_outputs = [ 
         s_trial_index, s_run_no, s_user_logs, s_current_trial_data, s_user_session_id,
-        s_current_run_image_list, s_num_trials_this_run,
-        *outputs_ui_components_definition 
+        s_current_run_image_list, s_num_trials_this_run, *outputs_ui_components_definition 
     ]
 
-    # --- 事件绑定 ---
     btn_agree_and_start.click(
-        fn=handle_agree_and_start, # 新的处理函数
-        inputs=[user_name_input, user_gender_input, user_age_input, user_education_input], # 添加 request
-        outputs=[
-            s_user_session_id,          # 更新用户ID状态
-            s_show_experiment_ui,       # 更新控制界面可见性的状态
-            welcome_container,          # 更新欢迎页容器的可见性
-            experiment_container,       # 更新实验页容器的可见性
-            welcome_error_msg           # 用于显示欢迎页的验证错误信息
-        ]
-    )
-
-    btn_start.click(
-        fn=partial(process_experiment_step, action_type="start_experiment"),
-        inputs=click_inputs_base, # 移除了 gr.Request()
-        outputs=event_outputs,
-        queue=True # 保持队列
-    )
-    btn_left.click(
-        fn=partial(process_experiment_step, action_type="record_choice", choice_value="left"),
-        inputs=click_inputs_base,
-        outputs=event_outputs,
-        queue=True
-    )
-    btn_right.click(
-        fn=partial(process_experiment_step, action_type="record_choice", choice_value="right"),
-        inputs=click_inputs_base,
-        outputs=event_outputs,
-        queue=True
+        fn=handle_agree_and_start, 
+        inputs=[user_name_input, user_gender_input, user_age_input, user_education_input, gr.Request()],
+        outputs=[s_user_session_id, s_show_experiment_ui, welcome_container, experiment_container, welcome_error_msg]
     )
+    btn_start.click(fn=partial(process_experiment_step, action_type="start_experiment"), inputs=click_inputs_base, outputs=event_outputs, queue=True)
+    btn_left.click(fn=partial(process_experiment_step, action_type="record_choice", choice_value="left"), inputs=click_inputs_base, outputs=event_outputs, queue=True)
+    btn_right.click(fn=partial(process_experiment_step, action_type="record_choice", choice_value="right"), inputs=click_inputs_base, outputs=event_outputs, queue=True)
 
-# ==== 程序入口 ====
 if __name__ == "__main__":
-    # ... (与上一版相同的启动检查和打印逻辑) ...
-    if not master_image_list:
-        print("\n关键错误：程序无法启动，因为 TARGET_DIR 中没有找到图片。"); exit()
+    if not master_image_list: print("\n关键错误：程序无法启动，因无目标图片。"); exit()
     else:
-        print(f"从 '{TARGET_DIR}' 加载了 {len(master_image_list)} 张可用的目标图片。每轮实验将尝试选取 {NUM_TRIALS_PER_RUN} 张。")
-        if not METHOD_ROOTS: print(f"警告: 在 '{BASE_IMAGE_DIR}' 目录下没有找到候选方法子目录 (除了 '{TARGET_DIR_BASENAME}')。")
-        else: print(f"已识别的方法根目录: {METHOD_ROOTS}")
-        if not SUBJECTS: print("警告: SUBJECTS 列表为空。将无法从子目录加载候选图片。")
-        else: print(f"实验将针对以下 subjects (子目录名): {SUBJECTS}")
-        print(f"日志将尝试保存到 Hugging Face Dataset: '{DATASET_REPO_ID}'")
-        if BATCH_LOG_FOLDER: print(f"  - 批量运行日志 (如果启用): '{BATCH_LOG_FOLDER}/'")
-        if INDIVIDUAL_LOGS_FOLDER: print(f"  - 单个选择日志: '{INDIVIDUAL_LOGS_FOLDER}/'")
-        if not os.getenv("HF_TOKEN"): print("警告: 环境变量 HF_TOKEN 未设置。日志将无法保存到 Hugging Face Dataset。\n      请在 Space Secrets 中设置 HF_TOKEN。")
-        else: print("环境变量 HF_TOKEN 已找到。")
+        print(f"从 '{TARGET_DIR}' 加载 {len(master_image_list)} 张目标图片。每轮选 {NUM_TRIALS_PER_RUN} 张。")
+        if not METHOD_ROOTS: print(f"警告: '{BASE_IMAGE_DIR}' 无候选方法子目录。")
+        else: print(f"方法根目录: {METHOD_ROOTS}")
+        if not SUBJECTS: print("警告: SUBJECTS 列表为空。")
+        else: print(f"Subjects: {SUBJECTS}")
+        print(f"日志保存到 Dataset: '{DATASET_REPO_ID}'")
+        if BATCH_LOG_FOLDER: print(f"  - 批量日志文件夹: '{BATCH_LOG_FOLDER}/'") # 使用 BATCH_LOG_FOLDER
+        if INDIVIDUAL_LOGS_FOLDER: print(f"  - 单个选择日志文件夹: '{INDIVIDUAL_LOGS_FOLDER}/'")
+        if not os.getenv("HF_TOKEN"): print("警告: HF_TOKEN 未设置。日志无法保存。\n      请在 Space Secrets 中设置 HF_TOKEN。")
+        else: print("HF_TOKEN 已找到。")
         path_to_allow_serving_from = BASE_IMAGE_DIR
         allowed_paths_list = []
         if os.path.exists(path_to_allow_serving_from) and os.path.isdir(path_to_allow_serving_from):
             allowed_paths_list.append(os.path.abspath(path_to_allow_serving_from))
-            print(f"Gradio `demo.launch()` 将配置 allowed_paths: {allowed_paths_list}")
-        else: print(f"关键警告：配置的图片基础目录 '{path_to_allow_serving_from}' ({os.path.abspath(path_to_allow_serving_from) if path_to_allow_serving_from else 'N/A'}) 不存在或不是一个目录。")
-        print("启动 Gradio 应用程序...")
+            print(f"Gradio `demo.launch()` 配置 allowed_paths: {allowed_paths_list}")
+        else: print(f"关键警告：图片基础目录 '{path_to_allow_serving_from}' 不存在或非目录。")
+        print("启动 Gradio 应用...")
         if allowed_paths_list: demo.launch(allowed_paths=allowed_paths_list)
         else: demo.launch()